Alternate raft pathways cooperate to mediate slow diffusion and efficient uptake of a sphingolipid tracer to degradative and recycling compartments

The discovery that many endocytic cargoes enter cells through clathrin-independent pathways has triggered interest in the identification of the endocytic machinery and membrane-trafficking itineraries of these pathways. Among these non-clathrin pathways, some are based on a constitutive internalization mechanism, whereas others are initiated by specific signals, e.g. cytokine signaling (Lamaze et al., 2001; Mayor and Pagano, 2007). However, it has been documented that some cargoes might be endocytosed through alternate or sequential pathways, dependent on cell type and treatment (Deinhardt et al., 2006; Massol et al., 2004; Sandvig et al., 2004). For example, in cells that lack clathrin or caveolae, internalization can still occur via alternative pathways. Several bacterial toxins, as well as the amyloid precursor protein (App), combine raft-mediated association at the plasma membrane with subsequent endocytosis, assisted by components common to clathrin- and/or dynamin-mediated uptake (Abrami et al., 2003; Schneider et al., 2008).

As well as determining which cargos are taken up by particular raft-associated mechanisms, it is important to understand the post-internalization trafficking routes following these entry mechanisms. It has been noted that various raft-borne endocytic cargoes are transported to different intracellular destinations. Some examples are enterotoxins such as cholera and shiga toxins, which are transported rapidly from the plasma membrane to the Golgi, and thereafter retrogradely to the endoplasmic reticulum, where toxic effects ensue (Lencer and Saslowsky, 2005). One well-studied class of raft-borne endocytic cargos are the glycosyl-phosphatidylinositol-associated proteins (GPI-APs) (Mayor and Riezman, 2004), which are thought to internalize via cdc42, into a specialized `GEEC' compartment (Sabaranjak et al., 2002) and thereafter most probably traffic to recycling endosomes and possibly Golgi (Nichols et al., 2001). Raft-associated adaptor proteins flotillin-1 and flotillin-2, which constitute a separate internalization pathway from either caveolin or GPI-AP uptake, target partially to lysosomes (Glebov et al., 2006; Langhorst et al., 2008).

Clathrin-independent internalization phenomena are associated with plasma membrane microdomains or lipid rafts, defined functionally by their structural requirement for sphingolipids and cholesterol (Hancock, 2006). The role of cholesterol in clathrin-independent endocytosis, such as caveola formation (Hooper, 1999), is considered diagnostic for the presence of raft-mediated mechanisms. A critical level of cholesterol is required for the organization of raft-borne molecules into nanoclusters at the cell surface, estimated to be ∼20-40% of the population at any given time (Goswami et al., 2008; Lasserre et al., 2008; Sharma et al., 2004). The cholesterol content in membranes has been shown to influence the integrity and functions of lipid rafts in signal transduction (Gomez-Mouton et al., 2004; Jury et al., 2007). It is also known that overload of cholesterol and sphingolipids influences the membrane trafficking of raft adaptor proteins (Pol et al., 2005) and lipid raft constituents, as occurs in the lipid storage diseases (Pagano et al., 2000; Simons and Gruenberg, 2000).

It is obvious from all of this data that cholesterol and sphingolipids are important for the uptake and trafficking of raft-associated molecules, but the rules governing the endocytic trafficking routes of these cargoes have not yet emerged. Although distinct intracellular pathways have been noted for certain raft-mediated cargo (Le and Nabi, 2003; Payne et al., 2007; Sabharanjak et al., 2002) we do not have a clear picture of how or whether the specific machinery of uptake also affects subsequent intracellular targeting. Neither has there been any observation as yet that would link the endocytic machinery (apart from cholesterol and sphingolipids) with a particular diffusion behavior (Goswami et al., 2008; Umemura et al., 2008).

In this study, we investigate the relationship between uptake mechanism, surface behavior and endocytic targeting of the recently characterized sphingolipid and raft-interacting probe, sphingolipid binding domain (SBD), in mammalian neuroblastoma cells. SBD, a fluorescently coupled peptide consisting of the V3-loop domain of the amyloid beta peptide (Aβ) (Fantini et al., 2002), was shown to interact with a subset of glycosphingolipids, sphingomyelin and cholesterol in artificial membranes as well as with cholesterol- and sphingolipid-dependent microdomains in neurons (Hebbar et al., 2008; Steinert et al., 2008). Here, we show that SBD uptake is dependent on several different mechanisms – flotillin, cdc42, and dynamin – that have generally been viewed as independent and parallel pathways (Glebov et al., 2006; Mayor and Pagano, 2007; Sabharanjak et al., 2002). Suggestive of synergism between these mechanisms, interference with individual components only partially suppresses uptake of SBD into cells, whereas knocking out the flotillin and cdc42 pathways simultaneously nearly eliminates uptake. By contrast, drastic changes in the raft-like diffusion behavior of SBD are seen when either or both of these uptake pathways are perturbed. We also observe that knockout of either cdc42- or flotillin-mediated uptake leads to a trafficking switch, as do conditions of altered cholesterol storage. This suggests that these alternate pathways favor different subcellular itineraries that are related to cholesterol-induced changes in the intracellular targeting of sphingolipids.

The fluorescently tagged SBD (Mahfoud et al., 2002) peptide acts as a sphingolipid-interacting tracer with potential applications in the study of neurodegenerative disease (Hebbar et al., 2008; Steinert et al., 2008). For this reason, we examined the uptake characteristics and intracellular trafficking of the SBD marker in neuroblastoma SH-SY5Y cells.

We determined the kinetics of cellular internalization of SBD in SH-SY5Y cells (Fig. 1A), by normalizing the unit fluorescence intensity of internalized label at a given time point against the average unit fluorescence after complete internalization at 60 minutes. Tetramethylrhodamine (TMR)- or Oregon green (OG)-labeled SBD was added to cells at a concentration of 10 μM and the cellular fluorescence quantitatively assessed at different time points. SBD is internalized rapidly, with >80% of maximum fluorescence obtained after 15 minutes. The internalization kinetics were nearly identical between TMR- and OG-tagged SBD (Fig. 1A). After ∼15 minutes of incubation, internalization of SBD-TMR and SBD-OG gave rise to discrete labeled compartments in the cytoplasm (Fig. 1B). The identity of these compartments was investigated, as described below.

In order to determine the uptake and intracellular trafficking route of SBD in neuroblastomas, we compared its localization to commonly used endocytic markers at early time points. First, the colocalization of SBD with the known GM1-binding marker Cholera toxin B (CtxB) and the clathrin-dependent marker transferrin (Tfn)-Alexa594 was re-examined (Fig. 2A-E) under different labeling conditions to those previously used (Steinert et al., 2008). Here, loading of the cells with fluorescent label was carried out for shorter time periods (10 or 15 minutes) and on ice, in order to ensure synchronous uptake upon warming to 37°C, starting at time 0. CtxB labelings were carried out in the absence of antibody cross-linking, because this procedure could possibly affect uptake and subsequent trafficking behavior. Quantification of colocalization was determined by the Costes thresholding method, and reported as the thresholded Manders' coefficient for the SBD channel (tM^SBD; see Materials and Methods and Fig. 2E).

Due to its association with ganglioside, CtxB is expected to localize at least partially to sphingolipid-rich microdomains (Shogomori and Futerman, 2001; Wolf et al., 1998). In spite of this, SBD and CtxB colocalized very little in early endocytic vesicles and sorting compartments (1 and 15 minute time points), comparable to the colocalization of SBD with Tfn (Fig. 2A-C,E). This closely matched previous results (Steinert et al., 2008) and confirmed that these two sphingolipid-dependent markers follow distinct intracellular routes. To confirm that the low degree of colocalization was not due to CtxB following a predominantly clathrin-dependent pathway in this cell type, colocalization between CtxB and Tfn was also checked, but was found to be very low, as expected (Fig. 2D,E).

Flotillins are multiply acylated, membrane-associated scaffolding proteins involved in the assembly of non-clathrin and non-caveolar endocytic platforms of ∼50-100 nm (Stuermer et al., 2001) that are largely non-overlapping with domains carrying CtxB or GPI-AP (Glebov et al., 2006; Langhorst et al., 2008). Flotillin-2 forms hetero- and homotetrameric complexes with flotillin-1, and is required for the maintenance of flotillin-1, but has a slightly different distribution from flotillin-1 (Langhorst et al., 2005; Solis et al., 2007). A flotillin-2–GFP (GFP, green fluorescent protein) transfected fusion protein (Neumann-Giesen et al., 2004; Steinert et al., 2008) and endogenous flotillin-2 were compared to SBD as a second raft-associated marker. After 1 minute, SBD showed very little colocalization with flotillin-2–GFP (presumably because flotillin-2–GFP is systemically expressed, whereas SBD is added externally), but this rose to ∼40% after only 10 minutes, similar to its degree of colocalization with the endogenous flotillin-2 protein (Fig. 2F-I). Again, this confirmed earlier results with non-synchronized uptake.

To test whether the SBD is endocytosed through a similar pathway to that of the GPI-APs, which are localized partially to nanoclustered, raft-dependent domains and endocytosed through a specialized GPI-AP-enriched early endosomal compartment (GEEC) (Sabharanjak et al., 2002), we expressed cyan fluorescent protein (CFP)-tagged GPI-AP protein (GPI-AP–CFP) in SH-SY5Y cells (Keller et al., 2001). Upon addition, SBD is not coincident with GPI-AP–CFP but rapidly colocalizes (10 minutes; Fig. 2J-L), showing that SBD probably enters the GEEC pathway.

The Rho family small GTPase cdc42 mediates the endocytosis of GPI-APs via the GEEC pathway, and this process does not involve dynamin (Sabharanjak et al., 2002). To test whether cdc42 also has a role in SBD endocytosis, we expressed wild-type and dominant-negative (DN, cdc42^N17) forms of GFP-tagged cdc42 in SH-SY5Y cells, and checked for both colocalization and a change in the uptake of SBD. SBD colocalized to a large extent (∼50%), mainly at the plasma membrane, with wild-type cdc42-GFP after 15 minutes (Fig. 2M,O). However, in cells expressing the dominant-negative form (cdc42-DN–GFP), the colocalization with SBD was dramatically reduced (Fig. 2N,O), as well as its uptake (see following section).

Dynamin is a large GTPase that mediates many forms of endocytosis and vesicle formation through its ability to tubulate and sever membranes (Praefcke and McMahon, 2004). Because dynamin-mediated uptake is thought to be functionally separate from the small GTPase-mediated mechanisms such as RhoA and cdc42 that internalize certain raft cargoes (Glebov et al., 2006; Mayor and Pagano, 2007; Payne et al., 2007; Sabharanjak et al., 2002), and because cdc42-DN only partially inhibited uptake (∼35% inhibition; see below), we asked whether dynamin might be involved in a parallel uptake mechanism for SBD. In order to test the role of dynamin in the internalization of SBD, we used two approaches. First, SH-SY5Y cells were treated with Dynasore (Sigma), a cell-permeable drug that inhibits dynamin-1 and dynamin-2 (Macia et al., 2006). Dynasore treatment, as expected, almost completely blocked the uptake of the control Tfn-Alexa594 (Fig. 3A), whose clathrin-mediated internalization requires dynamin (Sever et al., 2000). To our surprise, Dynasore also inhibited internalization of SBD, the uptake being 46% that of untreated control cells (Fig. 3A,B).

As an alternative method to investigate the dynamin-dependence of SBD uptake, we transfected SH-SY5Y cells with a dynamin-2 dominant-negative mutated version, dynK44A DN (Damke et al., 2001) that suppresses the function of the endogenous dyanmin-2 protein. Whereas the amount of Tfn that entered cells transfected with dynamin-2 WT plasmid was not affected, the entry of Tfn in cells transfected with dynK44A DN was very low (Fig. 3B,C). Similarly, SBD internalization was reduced by more than half in cells transfected with dynK44A DN compared with cells transfected with dynamin-2 WT (Fig. 3B,C). We conclude that the internalization of SBD is not only mediated by cdc42 but also by dynamin.

To confirm that clathrin is not involved in the internalization of SBD, we tested the uptake of SBD in the presence of drugs that inhibit clathrin-mediated endocytosis. To inhibit clathrin-mediated endocytosis, SH-SY5Y cells were treated with 5 μg/ml chlorpromazine (CPZ) for 30 minutes prior to the addition of SBD. This treatment, as expected, completely blocked the endocytosis of Tfn (Fig. 3D,E), which requires clathrin-mediated endocytosis. In comparison, chlorpromazine did not inhibit the entry of SBD (Fig. 3D,E).

Neuronal cells, including SH-SY5Y, do not express caveolin (Gorodinsky and Harris, 1995). The efficient internalization of SBD by SH-SY5Y cells and other neurons (Steinert et al., 2008) implies that the uptake of SBD is independent of caveolin.

As SBD showed colocalization with flotillin-2 after endocytosis (see Fig. 2), and because the internalization of both Aβ and the App precursor were shown to depend on flotillin as well as dynamin (Saavedra et al., 2007; Schneider et al., 2008), we next investigated the role of flotillin-2 in mediating SBD uptake and intracellular transport. We used siRNA to knock down the expression of flotillin-2 (see Materials and Methods), which we found reduced the amount of flotillin-2 expressed in cells to ∼20% of normal (supplementary material Fig. S1) This was similar to the findings of other authors (Schneider et al., 2008). It was reported previously that loss of flotillin-2 also leads to proteasomal degradation of flotillin-1, reducing the possibility that loss of flotillin-2 could be masked by the remaining flotillin-1 (Langhorst et al., 2005; Solis et al., 2007). Knockdown of flotillin-2 reduced SBD uptake to ∼69% of normal (Fig. 4A).

Clostridium toxin B was used to inhibit the activity of Rho family GTPases, including cdc42, which had already been implicated in toxin SBD uptake because of the inhibition by cdc42-DN–GFP. After treatment with clostridium toxin, SBD uptake was again inhibited, moderately, to ∼74% of normal levels (Fig. 4A).

Inhibition of SBD uptake was determined by calculating the total internalized fluorescence per pixel within cells, excluding the plasma membrane, after imaging 60 minutes post-labeling (see Materials and Methods). When the two treatments were combined, flotillin-2 siRNA and inhibition with clostridium toxin together gave effects on SBD uptake that were much stronger than additive (reduction to 9% of normal) (Fig. 4A). Tfn uptake was, by contrast, unaffected by the combined treatment, indicating that the drastically reduced uptake of SBD was not simply due to cellular damage.

Because the foregoing method of assaying the amount of SBD taken up by the cells is restricted to quantification of SBD that has been internalized and is no longer associated with the plasma membrane, we decided to examine the effects of flotillin-2 siRNA and clostridium treatments on uptake and association of SBD by using fluorescence activated cell sorting (FACS). The results of this experiment paralleled those of the previous image analysis uptake assay, in that SBD appeared to be taken up less efficiently after either treatment alone, but that both treatments together had the most drastic effects (Fig. 4B). Using this assay we were able to assess association with the membrane, as opposed to uptake per se, by comparing total cell fluorescence pre- and post-washing. This is indicated by the difference between the red and the green traces, which show pre-washing and post-washing total fluorescence per cell, respectively (plotted in Fig. 4B).

By this measure, the amount of SBD loosely associated with the cells appeared to be reduced after either or both of the treatments, because the difference between pre- and post-washing was substantially smaller. Indeed, even the absolute levels of pre-washing fluorescence, as indicated by the red traces, were reduced by siRNA flotillin-2 and clostridium toxin, suggesting less initial association of SBD with the membrane. Uptake was also reduced by these treatments, shown by a decrease in absolute levels of post-washing fluorescence (green traces). After both treatments (last bar in lower graph, Fig. 4B), association with the membrane and uptake were both strongly abrogated, as both red and green traces pre- and post-washing were close to control (unlabeled) fluorescence levels. From these results, we conclude that the activity of either of these two molecules (flotillin-2 or a Rho GTPase) is required to allow some association and uptake of SBD at the membrane; however, both flotillin-based and Rho-GTPase-based mechanisms together synergize to effect efficient and complete SBD uptake.

Using fluorescence correlation spectroscopy (FCS), we tested the effects of blocking different mechanisms of SBD uptake on its diffusion behavior at the plasma membrane surface of neuroblastoma cells. Diffusion time of the fluorescent probe across a stationary confocal volume can be derived from the autocorrelation function G(τ) of fluorescent signals recorded from this volume over time. Different models that take into account one or more diffusing particles and two- or three-dimensional diffusion can be fitted to the G(τ) function. Because diffusion speed is a reflection of membrane fluidity, which is known to be decreased in raft domains (Bacia et al., 2004; Eggeling et al., 2008; Goswami et al., 2008; Pinaud et al., 2009), diffusion times (τ_D) from a confocal spot centered at the plasma membrane can be used to deduce membrane microdomain localization. This method was used previously to show that SBD displays a slow diffusion component at the plasma membrane, similarly to CtxB, but unlike uniformly distributed lipophilic dyes (Hebbar et al., 2008).

The fits to the correlation functions of fluorescent particles diffusing through the confocal volume are determined in multiple 20-second cumulative measurements, and are generally separable into two components, fraction 1 and fraction 2 (Frn1 and Frn2), from which diffusion times can be derived. τ_D1 (in the millisecond range) and τ_D2 (in the microsecond range) are the slow and the fast components, respectively, and correspond to 2D-diffusing, membrane-bound label (τ_D1), and to 3D-diffusing, unbound label (τ_D2). By comparison with control G(τ) values of unbound label in solution (with an average τ_D in the microsecond range), it can be inferred that τ_D2 is mostly attributable to this unbound population (Wohland et al., 2001). It should be noted that FCS measurements inside the cell also give a predominantly τ_D2 (microsecond) component, possibly due to internalized SBD that might diffuse through endosomes.

By comparing the proportion of particles that fall into the τ_D1 or τ_D2 categories (Frnτ_D1 or Frnτ_D2) it is possible to determine whether the binding of SBD is affected under different treatments. Because the proportion of τ_D1 and τ_D2 will be affected if the confocal volume is not centered on the plasma membrane, care was taken to ensure that the confocal volume was always properly positioned at the plasma membrane. Frnτ_D1 and Frnτ_D2 were plotted in control, untreated cells and in cells treated with either flotillin-2 siRNA or clostridium toxin, or both (Fig. 4C). The fraction of τ_D1 (the presumptive membrane-bound, millisecond range) fell from 44% in controls to 33% with flotillin-2 siRNA, 29% with clostridium treatment, and 19% with both treatments. Thus, the magnitude of the effect of knocking out either of the uptake mechanisms alone was similar, and removing both gave an almost perfectly additive effect on the binding of SBD. This implies that the effects of flotillin and cdc42 on the binding of SBD are additive, if one assumes that a decreased proportion of Frn1 reflects a loss of binding. This is in contrast to the effects on uptake, which we saw in the previous section were synergistic.

By plotting the percentage fraction of τ_D1 in the different readings as a histogram instead of an average (Fig. 4D), it can be seen that the distribution of readings shifts from predominantly τ_D1 (>50% putative membrane-bound particles) to smaller fractions of τ_D1 (20-30%) under either siRNA treatment or treatment with clostridium toxin, and even smaller fractions (<10%) under both treatments simultaneously.

A previous study (Hebbar et al., 2008) that characterized the diffusion behavior of SBD on neuroblastoma cells showed that the diffusion rate of SBD averages ∼50 milliseconds (also confirmed here; Fig. 5A) but, when plotted as a histogram, the diffusion rate distributes into two separate categories, one of ∼1-10 milliseconds and the other of >30 milliseconds (see Fig. 5B). In that study, CtxB, but not the non-raft markers, also showed a bimodal distribution like that of SBD, and the slow component could be eliminated by methyl-β-cyclodextrin (mβCD). This suggested that the long diffusion time might correspond to the proportion of molecules that are associated (transiently) with a raft or nanocluster.

We now wanted to know whether the bimodal diffusion distribution was related to the different uptake mechanisms that affect SBD internalization into cells. Indeed, when treated with either clostridium toxin or flotillin-2 siRNA, or both, the slow component disappeared, being reflected in the much faster average τ_D (Fig. 5A) and in the shift of the histogram distribution to between 1-10 millisecond (Fig. 5B) under each of the three treatments.

In contrast to the τ_D1 values, τ_D2 values were not as strongly affected, although the average τ_D2 sped up by about 50% (Fig. 5C), and the histogram of τ_D2 did show a shift of τ_D2 towards faster values under the combined flotillin-2-RNAi–clostridium treatment (but not either treatment alone) (Fig. 5D). This shift in τ_D2 could be explained if τ_D2 derives from freely diffusing SBD both outside the cell and inside the cell, because extracellular SBD has an average τ_D of ∼115±5.8 microseconds (exclusively <200 microseconds; n=24 readings) whereas cytosolic τ_D averages 499.3±35.7 microseconds (exclusively >200 microseconds; n=45 readings) (supplementary material Fig. S2). Less efficient binding would also result in less SBD entering the cell, and therefore faster average τ_D2 values.

In order to determine whether the decrease in τ_D values was a result of a general increase in membrane fluidity induced by the RNAi and clostridium treatments, we measured the τ_D values of a uniformly distributed lipophilic membrane dye, dialkylindocarbocyanine-C18 (DiI). The average values and histogram distribution of τ_D remained unchanged after individual and combined treatments (supplementary material Fig. S3).

The results above suggest that the combined effects of flotillin and Rho family GTPase on binding of SBD are stronger than either one alone and are additive; however, because removing either flotillin or Rho GTPases completely abolishes the slow diffusion component of SBD, both of these mechanisms must be absolutely required for this raft-like diffusion behavior.

To test the involvement of Rho family GTPases and flotillin in regulating SBD intracellular targeting, colocalization studies were carried out with SBD and intracellular markers after interference with these pathways, using the same methods as above. First, flotillin-2–GFP localization was confirmed in SH-SY5Y neuroblastomas to localize ∼50% to an endolysosomal compartment identified by lysotracker red (supplementary material Fig. S4), in agreement with earlier studies (Langhorst et al., 2008). In control cells, SBD efficiently targeted early endosomes at 30 minutes after endocytosis, where it colocalized to ∼50% with early endosomal antigen 1 (EEA1) and this was unaffected by interference with clathrin-mediated uptake using chlorpromazine (Fig. 6A,B). siRNA knockdown of flotillin-2 significantly reduced the degree of colocalization of SBD with EEA1 (Fig. 6C,F), indicating that flotillin internalization is a prerequisite for further delivery of SBD to an early endosomal sorting compartment.

Next, we tested the role of flotillin-2 in mediating SBD trafficking to recycling endosomes and lysosomes, respectively. In control cells, SBD localized ∼50% in rab11-GFP-positive recycling endosomes, and ∼45% in acidic lysotracker-positive lysosomes after 1 hour (Fig. 6D-F). This late association with recycling endosomes is in accordance with earlier characterization in another cell type (Steinert et al., 2008), showing that SBD takes a slow route through sorting endosomes before it reaches recycling and degradative compartments. Notably, knocking down expression of flotillin-2 did not affect the targeting of SBD to rab11-positive recycling endosomes at all, in spite of its effect on EEA1 (Fig. 6D,F; above). By contrast, and similarly to its effect on early endosomes, flotillin-2 siRNA strongly decreased the amount of SBD trafficking to lysosomes (Fig. 6E,F).

We next sought to identify the intracellular trafficking pathways regulated by cdc42-mediated uptake of SBD, because we found in the previous experiments that uptake is mediated by a combination of flotillin and a Rho GTPase, most likely cdc42. Using clostridium B toxin to inhibit the activity of cdc42 (along with other Rho family GTPases), we found that the transport of SBD to recycling endosomes was largely blocked (Fig. 7A,C). However, SBD trafficked normally to lysosomes at 1 hour after endocytosis in spite of the inhibition of Rho GTPases (Fig. 7B,C). Together, these data show that inactivation of Rho GTPases inhibits the SBD trafficking to recycling endosomes, but not to lysosomes. Conversely, knockout of flotillin-mediated uptake inhibits SBD trafficking to lysosomes, leaving the recycling pathway unaffected.

As described in the previous section for inhibitors of uptake, we asked whether cholesterol was involved in intracellular trafficking of SBD in SH-SY5Y neuroblastomas. We had earlier determined that cholesterol was required for uptake in both mammalian and insect cells (Hebbar et al., 2008) and that sterol content modulated the trafficking of SBD to a degradative pathway in insect cells (Steinert et al., 2008). We first depleted cells of cholesterol using mβCD and observed cholesterol removal by staining the cells with filipin, a fluorescent polyene antibiotic (Beknke et al., 1984). High concentrations of mβCD (10 mM) were seen previously to result in effective depletion of cholesterol (50% reduction) from SH-SY5Y cells and subsequently blocked the uptake of SBD (Hebbar et al., 2008). Similar results were seen here (supplementary material Fig. S5), but we found that the uptake of SBD was not eradicated by milder cholesterol depletion using only 2 mM mβCD (although at this concentration, the diffusion speed at the membrane is affected; M.M., T.W. and R.K., unpublished results). Using this treatment, we were therefore able to assess the effect of cholesterol depletion on SBD trafficking. SBD trafficking to recycling endosomes was inhibited by mild cholesterol depletion (Fig. 8A,B) to a similar extent as with cdc42 inhibition (i.e. from ∼50% to ∼20%; see Fig. 7). Repletion of cholesterol to the membrane with mβCD-cholesterol complexes restored trafficking to the recycling compartment (Fig. 8A,B). Cholesterol depletion also substantially inhibited trafficking of SBD to lysosomes, and this could again be restored by cholesterol repletion (Fig. 8C,D).

We next wanted to determine the effect of cholesterol in the endolysosomal system on SBD trafficking. To induce intracellular cholesterol storage, we used U18666A, a drug known to lead to accumulation of cholesterol in lysosomes, similar to the accumulation seen in Niemann-Pick type C and the lipid storage diseases (Roff et al., 1991). In U18666A-treated SH-SY5Y cells, cholesterol accumulated in enlarged lysosomes, as expected (supplementary material Fig. S4). In U18666A-treated cells, we observed that this disrupted cholesterol distribution resulted in markedly more SBD accumulation in lysosomes than in controls (Fig. 8D) compared to untreated cells (P<0.01). Depletion of cholesterol from U18666A-treated cells with mβCD in turn reduced colocalization of SBD with lysosomes to lower than normal levels (Fig. 8C,D). Taken together, these results suggest that the intracellular trafficking behavior of SBD is sensitive to cholesterol content in both the early and late parts of the pathway.

The uptake mechanisms and intracellular trafficking of SBD, a fluorescently tagged sphingolipid interacting peptide probe, were examined in a human neuroblastoma line, SH-SY5Y. Unlike other microdomain-associated cargoes that have been characterized thus far, SBD was found to be endocytosed approximately equally by a cdc42-mediated pathway, and by a second pathway involving the lipid-raft-associated adaptor protein, flotillin (Langhorst et al., 2005; Mayor and Pagano, 2007). cdc42 has been implicated in regulating the GEEC pathway, specific for the uptake of GPI-linked proteins (Sabharanjak et al., 2002), which in some cases can also be endocytosed by flotillin-mediated mechanisms, although this is disputed (Glebov et al., 2006; Langhorst et al., 2008). Accordingly, SBD colocalized extensively in endosomes containing cdc42-GFP, flotillin-2–GFP and GPI-AP–GFP, although it segregated almost completely from another raft- and sphingolipid-interacting marker, CtxB.

Because SBD uptake is strongly affected by dynamin interference, and because knockout of Rho-GTPases (which include cdc42) and flotillin together virtually eliminates SBD uptake, we have to assume that dynamin takes part in one or other of the uptake mechanisms (i.e. cdc42 and/or flotillin), and does not constitute part of a third independent mechanism. According to the studies of Mayor and colleagues (Chadda et al., 2007; Sabharanjak et al., 2002) with GPI-APs, the cdc42 pathway is non-overlapping with dynamin mechanisms. However, a scenario combining the two would not be unprecedented because interleukin 2 receptor (IL2-R) endocytosis, which is also not clathrin-mediated, is simultaneously dynamin- and Rho-dependent (Lamaze et al., 2001). Furthermore, uptake that is jointly mediated by dynamin and flotillin has been documented multiple times. For example, and particularly relevant for this study, Schneider et al. reported that flotillin is involved in the uptake and clustering of App, even though App also requires dynamin and clathrin-adaptor proteins (Schneider et al., 2008). Uptake of Aβ (the parent molecule of SBD) is also dynamin-dependent (Saavedra et al., 2007). Zhuang and colleagues also describe a dynamin- and flotillin-mediated mechanism (Payne et al., 2007) responsible for the uptake of cationic agents like polyethylenimine (PEI) and Lipofectamine, but this was unaffected by cholesterol depletion.

In this study, we used three different methods to assay the effects of interference with flotillin- and Rho-GTPase-mediated pathways on SBD binding and uptake. First, quantification of internalized SBD by standard fluorescence imaging showed that removing either the flotillin pathway or the Rho-GTPase pathway by treatment with clostridium toxin exerted moderate effects on SBD uptake, whereas knockout of both mechanisms together nearly eliminated uptake. FACS quantification indicated that this reduced uptake was associated with reduced cell-surface-bound (i.e. washable) SBD after either treatment, and with stronger loss of cell-surface binding after both treatments. Accordingly, the FCS experiments also showed that an increased fraction of SBD diffused at the fastest rate, indicative of less binding, but these effects appeared to be additive for each treatment. Because the effects on actual uptake are probably synergistic, not additive, the results argue for involvement of both mechanisms in a single or strongly interacting internalization process.

The notion of a single, highly efficient uptake process dependent on both flotillin and cdc42 simultaneously is also supported by our observation that knockout of either mechanism alone completely abolishes the slowest diffusion component (see below). Synergism between two mechanisms might also explain the situation noted by Glebov and colleagues (Glebov et al., 2006), in which simultaneous DynK44A DN and flotillin-siRNA knocked out CtxB uptake by 80%, even though the effect of DynK44A DN alone was much less severe, and flotillin-siRNA alone had no significant effect. Thus, on the same principle by which geneticists predict protein interactions based on synergistic interactions between different mutations, we predict that there might be a direct interaction between flotillins and cdc42.

In a previous study using FCS, we showed that diffusion of SBD at the plasma membrane distributed into two categories: those with medium τ_D values (∼1-10 milliseconds) and those with slow τ_D values (∼50 milliseconds), whereas the non-raft-associated lipid analogs DiI and 4,4-difluoro-4-bora-3a,4a-diaza-s-indacene (BODIPY)-Sphingomyelin diffused only in the medium (1-10 milliseconds) range. Interestingly, CtxB in that study showed a very similar bimodal distribution, as did another presumptive raft-localizing sterol lipid analog [M.M., T.W., Gary Jennings (JADO Technologies, GmbH, Dresden, Germany) and R.K., unpublished results]. Suggestively, the longer diffusion time could be eliminated by mβCD and sphingolipid depletion (Hebbar et al., 2008) (M.M., T.W. and R.K., unpublished results) (see also Lasserre et al., 2008), indicating that the slowly diffusing fraction might correspond to the proportion of molecules that are associated (transiently) with a raft or nanocluster (Sharma et al., 2004). Here, we wondered whether this apparently raft-diagnostic long diffusion time corresponded to uptake by cdc42- and/or flotillin-mediated mechanisms. Surprisingly, whereas neither of these mechanisms was absolutely required to achieve some uptake (see previous section), the slow diffusion component was completely abolished when either mechanism was knocked out. This suggested that the slow diffusion might in fact correspond to a highly efficient, cdc42+flotillin synergistic uptake mode. A model proposing a link between diffusion behavior and uptake mode is shown in Fig. 9. A number of approaches could be used to test this model, including colocalization by super-resolution fluorescence imaging methods, or fluorescence cross-correlation spectroscopy, in parallel with studies of uptake kinetics.

Similar bimodal diffusion behaviors to that displayed by SBD have been seen before, notably in a recent study by Pinaud et al. (Pinaud et al., 2009), and earlier by Lommerse and colleagues (Lommerse et al., 2006). Other authors have recently derived diffusion constants from anisotropy and fluorescence recovery after photobleaching (ARAP and FRAP) (Goswami et al., 2008) and high-speed single-particle tracking (Umemura et al., 2008) on raft- and non-raft molecules; however, these either showed complete immobility associated with nanoclusters, or uniformly distributed hop-diffusion behaviors, irrespective of raft or non-raft localization. By contrast, Lassere and colleagues did find slow diffusion of raft-linked molecules (Lasserre et al., 2008). Unlike the present study, however, the diffusion behaviors in these studies were not linked to a particular uptake mechanism per se, but rather to cytoskeletal organization and raft lipid content.

An important finding of this study was that the normally occurring combination of endocytic mechanisms, involving both cdc42 and flotillin, not only affected diffusion speed and efficiency of internalization, but also the subsequent trafficking route. After internalization, SBD is split roughly equally between recycling endosomes and lysosomes. Surprisingly, the balance between these two pathways appears to be strongly affected by the uptake mechanism. Targeting to recycling endosomes requires the activity of cdc42 (possibly implicating GEECs in this recycling pathway) because GEECs target mainly to recycling endosomes and this is mediated by cdc42 (Chadda et al., 2007). Moreover, SBD, like GEEC-associated cargoes, has a very low degree of colocalization with Tfn (Sabharanjak et al., 2002). By contrast, targeting through the degradative pathway had a strong requirement for flotillin. This switch might reflect a connection of the uptake machinery (i.e. flotillin) with the requirement for a degradative rather than biosynthetic pathway for sphingolipids, perhaps in association with cholesterol and sphingolipid content at the plasma membrane. Support for this view (summarized in Fig. 10) was given by cholesterol overload and depletion experiments (below).

Interestingly, we found that cholesterol as well as the uptake mechanism is involved in the regulation of the SBD intracellular pathway. Moderate depletion of cholesterol reduced trafficking of SBD to recycling endosomes as well as to endolysosomes. The suppressive effect of mβCD cholesterol depletion on SBD recycling might, however, be brought about indirectly because cdc42 activity and exchange at the membrane are sensitive to cholesterol levels (Chadda et al., 2007). By contrast, overloading the lysosomal pathway with cholesterol left recycling unaffected, but measurably increased the amount of SBD found in degradative compartments.

These results are consistent with the idea that SBD recognizes sphingolipid- and cholesterol-rich regions of the membrane, and might be following their pathway to degradation under conditions of excess cholesterol storage. In this scenario, high cholesterol conditions would induce the degradation of sphingolipid and/or cholesterol pools or their recycling back to the surface, rather than further metabolic processing and storage. Cholesterol depletion, by contrast, might shunt sphingolipids away from both the degradative and recycling pathways, because under low cholesterol availability, more lipid should be stored and/or processed, rather than cycled out or degraded.

We propose that the behavior of SBD trafficking might reflect the function of its parent molecule Aβ, albeit without the physiological consequences of Aβ administration. As such, SBD accumulation in certain cellular compartments could serve as a metabolic `stop storing fat' indicator, steered to these locations by the presence of high levels of sphingolipids and/or cholesterol (i.e. raft components), with the aid of accompanying endocytic accessory proteins such as cdc42 and flotillin.

SH-SY5Y neuroblastoma (ATCC, Manassas, VA) were cultivated at 37°C in Dulbecco's modified Eagle's medium (DMEM; Gibco) /F12 (1:1) supplemented with 10% fetal bovine serum (FBS; Gibco) and 1% antibiotic. The rat-flotillin-2–GFP cDNA was kindly provided by Sean Sweeney and subcloned into a mammalian expression vector pcDNA3.1 (Invitrogen) as described previously (Hebbar et al., 2008). GFP–Cdc42-WT and GFP–cdc42-DN were kind gifts of Sohail Ahmed. Plasmid DNA (1 μg) was used along with Lipofectamine 2000 (Invitrogen) for transient transfections according to manufacturer's instructions. Experiments were carried out approximately 48 hours following transfection.

mβCD, cholesterol, HEPES, PMSF, Chlorpromazine, and Dynasore were obtained from Sigma (St Louis, MO). Cholesterol depletion and overload was carried out as previously described (Hebbar et al., 2008; Steinert et al., 2008). Chlorpromazine (5 μg/ml) was used to inhibit clathrin-dependent endocytosis by pretreating for 30 minutes at 37°C before cell labeling. For inhibition of dynamin activity, 80 μM Dynasore was incubated with cells in serum-free medium at 37°C for 30 minutes prior to subsequent experiments. To inhibit Rho family GTPase activity, including cdc42, cells were incubated with Clostridium difficile toxin B (40 ng/ml; Calbiochem) at 37°C for 1 hour.

Cells were grown in eight-well chambers with 0.17 mm coverslip bottoms (Nunc, Denmark) one day before transfection. For transfection, 1 μl of Lipofectamine-2000 (Invitrogen) was diluted to 25 μl Opti-MEM I (Invitrogen), and the solution incubated for 5 minutes at room temperature. This mixture was then added to a second solution of siRNA (20 μM) in 25 μl Opti-MEM I and allowed to complex for 20 minutes at room temperature. After incubation, 150 μl of full growth medium without antibiotics was added to the Lipofectamine-2000–siRNA solution, and the resulting 200 μl was added to the cells. On day 2, cells were transfected with siRNA again according to the above procedure. In this second round of transfection with siRNA, cells were co-transfected with a GFP-tagged plasmid for use as a marker. At 48 hours after this co-transfection, cells were used for further experiments. A scrambled, non-specific siRNA construct was used as a control, and siRNA with the sequence 5′-UAACCUCCUCACUGAAGGdTdT-3′ was used against flotillin-2 (both from 1st BASE, Singapore).

Cells were seeded at a density of 10⁶ cells/ml into eight-well chambers with 0.17 mm coverslip bottoms (Nunc, Denmark). Experiments were conducted 24-72 hours post-seeding. Lysotracker staining of acidic compartments was done by incubating cells for 2 hours at 37°C with 75 nM lysotracker red (Invitrogen) in normal growth medium.

For labeling of free cholesterol with filipin, cells were fixed in 4% paraformaldehyde (PFA) for 20 minutes followed by washes in HBSS-HEPES. Cells were incubated with 50 μg/ml filipin (Sigma) for 30 minutes and then washed before imaging.

CtxB labelings were done with Vybrant Alexa594 (Invitrogen). Cells were incubated with 10 μg/ml CtxB in full growth medium. CtxB labeling was carried out without antibody cross-linking. Alexa-Fluor 594 Tfn (5 μg/ml; Molecular Probes) was added to cells in serum-free medium. For endocytic tracing experiments, cells were incubated with fluorescently labeled endocytic markers on ice for 15 minutes then washed to remove excess dyes before warming for the indicated times, and imaging. Double labelings with CtxB (Vybrant) or Tfn-Alexa and SBD were done by first incubating CtxB or Tfn for 10 minutes on ice, then washing quickly, and incubating with SBD (10 μM) for 15 minutes on ice, then washing and immediately imaging at 37°C.

Labelings for FCS experiments were carried out at a concentration of 10 nM for DiI and 30-50 nM for SBD.

Cells were fixed with 4% PFA for 20 minutes at room temperature and then blocked (2% BSA, 0.5% Triton-X-100 in PBS) for 15 minutes at room temperature. The primary antibody (1 mg/ml) was added to the cells at a 1:200 dilution and incubated for 1 hour at room temperature. Alexa488- or Alexa568-coupled secondary antibody (Invitrogen; 1 mg/ml) was added to the cells (1:1000 in 0.2% BSA, 0.05% Triton-X100 in PBS) for 30 minutes. Cells were washed three times with 0.2% BSA and 0.05% Triton-X-100 in PBS between each step. Anti-EEA1 and anti-flotillin-2 monoclonal antibodies were purchased from BD Biosciences.

For uptake experiments, images were acquired with a CoolsnapHQ CCD camera on a Deltavision (Applied Precision) inverted fluorescence microscope with a 60× 1.42NA oil lens (Olympus) and a standard (green ex490/20, em528/38; red ex555/28, em617/73) filter set (Chroma). Quantification of images was performed using MetaMorph. Internalized fluorescence was determined by drawing borders around individual cells and subtracting non-cellular background. Uptake was calculated as a percentage of control in non-treated cells after 60 minutes of post-labeling chase (fluorescence units per pixel in treated cells divided by fluorescence units per pixel in untreated cells, times 100). All photomicrographs in a given experiment were exposed and processed identically for a given fluorophore. The Costes thresholded colocalization algorithm was performed using ImageJ (rsb.info.nih.gov/ij) plugins `Colocalization Test' and `Colocalization Threshold' by T. Collins and W. Rasband, and `BG Subtraction from ROI' by M. Cammer and T. Collins (www.uhnresearch.ca/facilities/wcif/imagej). Colocalization was carried out as previously described (Steinert et al., 2008) and the thresholded Manders coefficient (tM) was given as a percent to express the degree of colocalization of the SBD channel with the other channel. Each data point consisted of two separate experiments, taking into account at least ten cells each.

SBD-TMR association with the plasma membrane and uptake by neuroblastoma SH-SY5Y cells was assayed by measuring the amount of SBD fluorescence associated with the cells by FACS measurement before and after washing. Briefly, untreated cells, or cells treated with clostridium toxin B or siRNA flotillin-2, or both, were labeled with SBD, as previously. SBD endocytosis was terminated by placing the cells on ice. The cells were immediately analyzed using a FACS BD flow cytometer in FL-2 channel without further washing. Alternatively, cells were extensively washed with cold PBS to deplete surface-bound SBD and then were analyzed using a flow cytometer in FL-2 channel. About 10,000 cells were analyzed for each sample in triplicate for each condition. The G-mean fluorescence intensity (GMFI) of the cells for each treatment was analyzed and compared with a negative control (unlabeled SH-SY5Y cells). The GMFI values of cells without washing derived from internalized SBD plus plasma-membrane-bound SBD. The GMFI values of cells with extensive washing were presumed to have derived only from internalized SBD. The plasma-membrane-associated SBD for each treatment was calculated by the difference in fluorescence (GMFI without wash minus GMFI with wash).