In cell fractionation studies in which ¹²⁵I-NGF is bound to PC12 cells in the cold, and the cells are washed and warmed to allow internalization of NGF-bound receptors [7], [35], [36], NGF caused rapid internalization of TrkA into endosomes that could be recovered in organelles that emerged when cells were mechanically permeabilized by a single passage through a tight passage created by a ball whose diameter is very close to that of a surrounding cylinder (Balch homogenizer [35], [36]). After 10 min, about 40% of the TrkA is internalized, compared to a background endocytosis of about 5% without NGF. Under these conditions, at least 30% of NGF was reproducibly associated with the detergent-insoluble pellet after extraction with 1% non-ionic detergent (Triton X-100, NP-40 or IGEPAL; see Table 1). In contrast, only 1–4% of ¹²⁵I-transferrin is associated with the detergent-insoluble pellet under identical experimental conditions (Table 1). The significant difference between the amount of NGF vs. transferrin associated with the detergent-insoluble pellet leads to the hypothesis that NGF receptors are recruited into DRMs that would float when the pellet was resuspended layered under an iodixanol equilibrium gradient.

We used a similar pulse-stimulation protocol to investigate the association of NGF and its receptors, TrkA and p75^NTR with DRMs: cells were bound to ¹²⁵I-NGF in the cold, then washed and warmed for defined periods. Cells were lysed in non-ionic detergent and the insoluble material was subjected to equilibrium flotation iodixanol gradients (Figure 1). The peak at ρ=1.155–1.165 g/ml is defined as DRMs, which separated from higher density non-floating material (Figure 1A). NGF was present in DRMs on the plasma membrane before warming (Figure 1A, 0 min) and persisted for 30 min. There was little change in the amount associated with the floating peak over time (Figure 1B), although the density of the floating peak increased transiently (Figure 1C). Rat dorsal root ganglia neurons displayed a similar floating DRM peak containing NGF, though the density of this peak was slightly higher than that derived from PC12 cells (Figure 1D). These data suggest that PC12 cells are a valid model for neurons for the study of the DRM fraction containing NGF receptors.

To test the hypothesis that lipid rafts play a role in sorting TrkA and p75^NTR into different endocytic pathways, we focused on the time points of 0 and 10 min. At these times, TrkA and p75^NTR associated with floating DRMs in a peak at the same density as ¹²⁵I-NGF (Figure 2; see below).

Previously, we showed that tubulin could be detected in the detergent-resistant pellet from PC12 cells [34]. Since tubulin can be palmitoylated and the palmitoyl group when attached to proteins often confers association with DRMs [37], [38], we asked whether tubulin could be detected in floating DRMs (Figure 2A). In previous work, in vitro reactions with ATP enhanced tubulin polymerization leading to increased amounts of microtubules in the detergent-resistant pellet [34]. These data show that in vitro reactions with ATP can be used to manipulate microtubule polymerization. In vitro reactions increased by 9-fold the amount of tubulin in the floating DRM peak (Figure 2). Under these conditions, NGF and TrkA both increased 4–5 fold (Figure 2). In contrast, p75^NTR was reduced by about half in the floating peak after in vitro reactions (Figure 2). Flotillin was not affected, indicating that in vitro reactions do not artifactually produce a general aggregation of membranes (Figure 2). Thus, the amount of NGF and TrkA receptors in floating DRM specifically correlates with the presence of microtubules.

We asked if treatments that are known to affect the amount and the activity of TrkA in lipid rafts also affect microtubules in DRMs. The ganglioside, GM1 has been shown to activate Trk receptors and prevent apoptosis in sympathetic neurons and PC12 cells, which is hypothesized to be due to increased TrkA association within lipid rafts, [39]–[43]. Overexpression of the enzyme that produces GM1, however, has also been shown to decrease amounts of specific proteins associated with rafts and suppress TrkA dimerization, which is required for signaling activity [44]. These data suggest that TrkA signal transduction causes its recruitment to lipid rafts. One possibility is that GM1 at very high levels may also dilute rafts or change the properties of the membrane such that signaling is impeded. To determine whether changes in lipid rafts affected the recruitment of TrkA and p75^NTR, we measured the effect of adding GM1 on the amount of NGF, its receptors, and microtubules in DRMs. GM1 increased NGF and Trk in DRMs more than 2-fold (Figure 3A, B). In contrast, p75^NTR and flotillin were affected by GM1 in the opposite way. GM1-treated cells had less than half the amounts of p75^NTR and flotillin in floating DRMs compared to those of control (Figure 3C). It is important to note that the fraction of p75^NTR and flotillin in DRMs is constitutively high, about 20% without GM1 treatment, compared to TrkA (∼2%). p75^NTR and flotillin are known to preferentially associate with lipid rafts in many different cell types, and this property may be related to their similar decrease in DRMs in GM1-treated cells. The data are consistent with high levels of GM1 diluting rafts, which affects proteins that preferentially or constitutively associate with rafts differently than proteins that transiently associate with rafts in response to stimulation.

We also found that the microtubules that associated with floating DRMs increased more than 3-fold after GM1 treatment (Figure 3A and C, tubulin). Thus, GM1’s effects, as with in vitro reactions that cause microtubules to polymerize, were to increase microtubules in DRMs, which correlated with increases in NGF and TrkA. In both cases p75^NTR behaved in the opposite manner. The data suggest that NGF is mostly bound to TrkA, not p75^NTR, in floating DRMs, because the changes in the distribution of NGF paralleled that of TrkA.

We used two different methods to break up the insoluble material in the detergent-resistant pellet: sonication and nuclease (Benzonase) treatment (see Materials and Methods). When sonication was used, the density of the floating peak was approximately 1.16 g/ml. Trk was associated with two peaks on these gradients, one of which coincided with the floating NGF peak (Figure 2). The other peak was of higher density (1.23 g/ml) and did not coincide with NGF (Figure 2). p75^NTR was also found in the 1.16 g/ml floating peak with NGF and Trk, and little was present in other regions of the gradient (Figure 2). A fraction of tubulin also remained in the non-floating bottom of the gradient under these conditions (Figure 2). These data indicate that a fraction of the microtubules in DRMs were reproducibly resistant to sonication. To further investigate this possibility we determined whether results could be obtained without sonication. We found that Benzonase treatment facilitated handling of the DRM fraction without sonication (Figure 4). Quantitative distribution of receptors into floating DRMs was similar after sonication or Benzonase treatment, but the receptors floated into a peak of slightly higher density (1.20 g/ml) and there was less non-floating material in benzonase-treated samples (Figure 4A, B). Actin filaments were also present in floating DRMs under these conditions (very few were detected after sonication), but there was no change in their association with DRMs after in vitro reactions (Figure 4C). Importantly, comparable increases in NGF and microtubules in floating DRMs after in vitro reactions were observed (Figure 4B,C).

TrkA was reproducibly dephosphorylated in floating DRMs. Under conditions where phospho-TrkA (pTrkA) was detected in the detergent-sensitive (P1M) fraction, and in endosomes (see below, Figure 7A), TrkA but not pTrkA was present in floating DRMs (Figure 4A). The presence of the tyrosine phosphatase, SHP-1 in floating DRMs (Figure 4A, B) suggests a mechanism by which TrkA is selectively dephosphorylated in this fraction.

The similar increases in NGF, TrkA and microtubules in DRMs in response to GM1 and in vitro reactions suggest that TrkA may bind to microtubules in this fraction. Indeed, TrkA was co-precipitated when microtubules were stabilized with taxol and immunoprecipitated from floating DRMs (Figure 5A). If biotinylated tubulin was added during the last 5 min of in vitro reactions, it was incorporated into floating DRMs, suggesting that newly polymerized microtubules were associated with this fraction (Figure 5B, biotin-tubulin). Streptavidin agarose beads recovered TrkA from this fraction and biotinylated tubulin was pulled down by TrkA immunoprecipitation (Figure 5B). p75^NTR was not reproducibly detected in microtubule immunoprecipitations in these experiments. These data suggest that microtubule polymerization and attachment to DRMs recruits TrkA.

Under these conditions, biotinylated tubulin accumulated in discrete foci at the plasma membrane of permeabilized cells (Figure 5C, D). Long, intact microtubules were not evident in the permeabilized cell preparation, so we cannot exclude the possibility of some kind of alternate assembly of tubulin subunits at these discrete foci on the plasma membrane. Nevertheless, our data above showing that NGF and TrkA are recruited into DRMs along with microtubules, but not p75^NTR or flotillin, indicate that sorting specificity is reconstituted.

We compared the effects of NGF on TrkA and p75^NTR in floating DRMs. Without in vitro reactions, NGF caused a 1.5- to 2-fold increase of both TrkA and p75^NTR in the floating peak (Figure 6A, no rxn). In contrast, after in vitro reactions, NGF caused TrkA to increase, and p75^NTR to decrease in the floating DRMs (Figure 6A, +in vitro reaction). It has been noted previously that NGF signaling enhances tubulin polymerization [45], [46]. These data, together with that showing that microtubules assemble and associate with floating DRMs during in vitro reactions, suggest that NGF signaling may enhance microtubule association with DRMs during these reactions. Indeed, NGF significantly increased amounts of microtubules in floating DRMs, which correlated with increased TrkA in this fraction (Figure 6B, TrkA and tubulin). In contrast, p75^NTR was significantly reduced in floating DRMs after NGF treatment under these conditions (Figure 6B, p75^NTR), and flotillin was unchanged by NGF (Figure 6B, flotillin). These results indicate that NGF differently affected localization of the two co-receptors in floating DRMs under conditions where microtubules assemble and associate with membranes.

One possible outcome of sorting in rafts could be for conveying receptors into different endosomes [5], [6], [7]. We asked whether TrkA could be detected in lipid rafts associated with microtubules in endosomes. We examined endosomes using organelle fractionation methods described previously [7], [34]. Organelles that emerged from mechanically permeabilized cells were subjected to velocity sedimentation followed by floatation equilibrium centrifugation on iodixanol gradients. Endosomes containing activated TrkA and p75^NTR were recovered from cells treated with NGF as previously described (Figure 7A) [7]. To isolate lipid rafts associated with endosomes, organelles released from mechanically permeabilized cells were treated with detergent and centrifuged at 100,000× g. The pellet was resuspended and applied to iodixanol velocity gradients that separate microtubules from other material as previously described [34]. TrkA was present in detergent-resistant endosomal fractions that contained microtubules, and amounts increased after NGF treatment and in vitro reactions that enhanced microtubule polymerization (Figure 7B, input, MT). When microtubules were immunoprecipitated from this fraction, TrkA was bound to them after NGF treatment and in vitro reactions (Figure 7B, +ATP, NGF, MTIP). In contrast, p75^NTR was barely detected in this detergent-resistant endosomal fraction, and none was bound to microtubules (Figure 7B, p75). Phosphorylated TrkA was not detected in the detergent-resistant endosome fraction, though it was present in endosomes (Figure 7A), which is consistent with it being dephosphorylated in DRMs extracted from whole cells (Figure 4). Since TrkA was phosphorylated in endosomes, we asked if the tyrosine phosphatase, SHP-1 was present. SHP-1 was detected only in trace amounts, or not at all, in endosomes; it was found in fractions at the bottom of equilibrium gradients, indicating that it was weakly associated and transiently bound to organelles during the first velocity gradient sedimentation (Figure 7A). The data suggest that a portion of TrkA is sorted into DRMs, dephosphorylated, and endocytosed by a mechanism that involves microtubules. p75^NTR does not employ this mechanism, though it associates with DRMs. Activated TrkA is endocytosed by a different mechanism that excludes the phosphatase, SHP-1 to form signaling endosomes (Figure 7A) [7].

Wild-type PC12 cells were obtained from Lloyd Greene (Columbia University) and grown on collagen-coated plates in RPMI 1640, 5% fetal calf serum, 10% horse serum as described [81]. ¹²⁵I-NGF was prepared as previously described [36]. In some experiments biotinylated lactoperoxidase was used (Sigma) and removed by binding to neuravidin beads (Pierce, Rockfork, IL) prior to separation of radiolabeled protein from free iodine.

PC12 cells (typically 0.5–1×10⁹) were harvested in PBS and washed in cold PEE (PBS/1mM EGTA/1mM EDTA), PGB (PBS/0.1% glucose/0.1% BSA) as described [35], [36]. For comparison of treatment conditions, equal volumes of cell suspension were dispensed. Where NGF was added, 1 nm NGF or ¹²⁵I-NGF was bound to a rotating cell suspension 1 h at 4°C in PGB. Unbound ligand was removed by a wash in PGB to avoid fluid-phase endocytosis. For GM1 treatments, cells were harvested, separated into two equal aliquots, and incubated in either serum-free media with 65 µM GM1, or serum-free media alone, for 5 hours at 37°C in a 5% CO₂ incubator. After this incubation period the cells were washed and ¹²⁵I-NGF was bound to the cells above.

Cells were fractionated directly or warmed in PGB to 37°C exactly 2, 10, or 30 min, followed by a temperature-quench in ice water. Cells were then centrifuged 100 × g for 3 minutes and washed with 5 ml PEE, followed by a wash with 5 ml buffer B (cytoplasmic ionic composition: 38 mM each of the K⁺ salts of aspartic acid, glutamic acid and gluconic acid, 20 mM MOPS pH 7.1 at 37°C, 10 mM potassium bicarbonate, 0.5 mM magnesium carbonate, 1 mM EDTA, 1 mM EGTA), and resuspended in buffer B with 5 mM reduced glutathione (abbreviated BB+G). Protease inhibitors were added to a final concentration of 174 µg/ml PMSF, 1 µg/ml o-phenathroline, 10 ng/ml pepstatin, 10 ng/ml chymostatin, 10 ng/ml leupeptin, and 10 ng/ml aprotinin. Cells were mechanically permeabilized by a single passage through a Balch homogenizer in buffer B as described [7], [35], [36].

Where in vitro reactions were performed, the permeabilized cell suspension was split and one sample was warmed for 15 minutes at 37°C with an ATP regenerating system (1 mM ATP, 8 mM creatine phosphate, 5 mg/ml [240 units/mg] creatine kinase). After reactions, the samples were quenched in ice water for 3–5 minutes.

Dorsal root ganglia neurons were obtained from 30–40 embryos at stage E13. Ganglia were dissected from embryos in Leibovitz’s L-15 media (Gibco BRL), washed Eagles Balanced Salt Solution, and treated with 0.05% trypsin for 25 min. Cells were centrifuged 200 × g 4 min, then resuspended and plated on polylysine/laminin-coated 10 cm dishes and cultured in MEM, 10% FBS, 0.2% glucose, 2 mM glutamine with antibiotics (Pen/Strep) in the presence of 1.7 nM NGF for 4–10 days. The yield was about 8 million cells total. To prepare DRMs from neurons, cells were rinsed in warm PBS, then cold PEE was added and plates placed on ice for 30 min. This caused the cells to lift from the dish without breaking apart. Cells were recovered by centrifugation bound to NGF, warmed, and DRMs were prepared as described above for PC12 cells, except that neurons were swollen in hypo-osmotic 0.1 × BB prior to mechanical permeabilization with the Balch homogenizer.

Large membranes were removed from the mechanically permeabilized cell suspension by centrifugation at 1000 × g [7], [35], [36]. Gradients used iodixanol mixed with buffer B plus glutathione (BB+G) as media; continuous gradients were prepared using a two-chamber mixer. The supernatant of the 1000 × g centrifugation (S1) was layered over 0–30% velocity gradients. For two dimensional separations (velocity followed by equilibrium gradients), five velocity gradient fractions were collected, mixed with 60% iodixanol to a concentration of 32.5% or greater, and overlaid with a continuous 0–30% iodixanol/BB+G gradient and centrifuged to equilibrium (16–18 hr). Refractive indices were measured using an Abbe refractometer (Bausch and Lomb) and converted to density using the formula ρ=R.I.× 3.4319–3.5851, which was determined empirically by weighing known concentrations of iodixanol in buffer B.

Cracked cell suspensions were centrifuged at 1,000 × g for 10 minutes (pellet=P1; supernatant=S1). The P1 pellet was resuspended in buffer B containing protease inhibitors. Triton X-100 or IGEPAL (Sigma) was added to a final concentration of 1% and the sample was vortexed and left on ice for at least 1 hour. The sample was centrifuged at 10,000xg for 10 minutes to separate the detergent-soluble membranes (supernatant=P1M) from the detergent-insoluble pellet. This pellet was then resuspended in 180 µl BB with protease inhibitors and Triton X-100 was again added to a 1% concentration. Iodixanol (OptiPrep^TM) was added to a 40% concentration and the samples sonicated for 2 × 5-second pulses. This step was necessary as in preliminary experiments without sonication, DNA and cytoskeletal elements often clogged the needle of the gradient fraction collector. Alternatively, instead of sonication, 2 µl Benzonase (Novagen, Madison, WI) was added to the resuspended detergent resistant pellet and the sample incubated on ice for 1 hour to break up DNA prior to adding iodixanol. The method used for each set of experiments presented is identified in the figure legends.

Samples were placed into 5 ml ultracentrifuge tubes and 10–40% (for sonicated samples) or 15–48% (for Benzonase-treated samples) continuous iodixanol gradients in BB were poured over the top of the samples at 4°C. Gradients were centrifuged to equilibrium at 100,000xg for 16–18 hours. Approximately 200 µl fractions were collected from the bottom of the ultracentrifuge tube. Radioactivity in each fraction was determined using a gamma counter. The refractive index of each gradient fraction was measured and the density calculated based on a formula determined empirically by weighing iodixanol/BB standards of known concentration. TCA was added to each fraction to a final concentration of 10%, and left overnight at 4°C to precipitate protein. Protein precipitates were recovered by centrifugation at 3,500 × g for 35 minutes or 10,000× g for 20 min, then washed in ice-cold acetone and re-centrifuged. Precipitates were air dried at room temperature and 7 M urea sample buffer (7 M urea, 125 mM Tris HCl pH 6.95, 0.1% w/v bromophenol blue with 100 mM DTT) was added and samples were heated to 55°C for 15 minutes before analysis by SDS-PAGE.

For immunoprecipitation of microtubules, 10 µM taxol was added to the resuspended detergent-resistant pellet (treated with Benzonase) and gradients to stabilize microtubules. Where indicated, 25 µg/ml biotinylated tubulin and 12.5 µM taxol (both from Cytoskeleton, Inc., Denver, CO) were added during the last 5 min of 15 min in vitro reactions prior to preparation of DRMs. For microscopy, the sample was brought to 10% glycerol, 5% BSA and 1∶100 Texas Red-streptavidin (Vector Laboratories, Burlingame, CA) were added and incubated on ice for 2 hr. The permeabilized cells were washed 3X in buffer B with 0.1% BSA and recovered by centrifugation at 100 × g, 3 min. The sample was resuspended in buffer B with 20% glycerol, mixed with VectaShield (Vector Laboratories) and viewed with a 100x objective on a Nikon E800 with Hamamatsu ORCA II detector. The presence of floating membranes was confirmed by western blots of gradient fractions in parallel gradients on one half of the sample. Fractions containing the floating membranes of density 1.21- 1.15 g/ml, determined by refractive index measurements, were pooled and buffer components were added to 10% glycerol, 1% BSA,150 mm M NaCl, 50 mM Tris pH 7.7, 1% IGEPAL, 1 mM EDTA, and 10 µM taxol. Where indicated, samples were incubated overnight at 4°C with anti-α-tubulin (DM1A, Sigma) or anti-TrkA (RTA, Upstate Biotechnology). Antibodies were recovered with Pierce Ultralink ProteinA/G beads; biotinylated microtubules were recovered with Neuravidin beads. Bead suspensions were rotated 1 hr at 4°C, then recovered by centrifugation 1000 × g for 5 min, washed twice in wash buffer (10% glycerol, 150 mm M NaCl, 50 mM Tris pH 7.7, 1% IGEPAL, 1 mM EDTA, 10 µM taxol), then once in 0.01X wash buffer, and resuspended in SDS-PAGE sample buffer.

Microtubules were immunoprecipitated from the detergent-resistant endosome fraction exactly as described [34]. Briefly, PC12 cells were treated with or without NGF and subjected to in vitro reactions as above. The organelles that emerged from mechanically permeabilized cells, which have been shown to contain TrkA bound to NGF [7], [35], [36], were incubated with 1% IGEPAL. Detergent-resistant material was concentrated by centrifugation at 100,000 × g, 1 hr, resuspended and applied to iodixanol velocity gradients. Fractions at the bottom of gradients containing microtubules, and control samples from the top of the gradient, were individually pooled and immunoprecipitated with anti-α-tubulin as described above except without taxol.

SDS-PAGE gels were run and western blotted to nylon-reinforced nitrocellulose (Schleicher and Schull, Dassel, Germany) as described [82]. Where different treatment conditions are compared in one experiment, all blots were incubated in the same antibody solution on the same day and exposed for the same amount of time. Blot incubations were performed in 5% nonfat dry milk, 150 mM NaCl, 50 mM Tris pH 7.7, 0.05% Tween 20, or conditions specified by the antibodies’ manufacturer. Secondary anti-mouse or –rabbit antibodies coupled to HRP (Amersham) were used and chemiluminescent signals generated by Amersham ECL^TM or Super Signal West Pico (Pierce). The blot was either exposed to X-ray film (Fuji medical x-ray film, HR-G 30) or exposed directly in a Fujifilm Intelligent Dark Box II with a cooled CCD camera (LAS-1000, or LAS-3000, Fuji Photo Film Co. Ltd, Japan). Blots were stripped of antibodies for re-probing with Restore (Pierce), TBS pH 2.0, or in 0.5 M NaCl, 0.2 M glycine, pH 2.8.

Chemiluminescent data captured directly or on film by the LAS-1000 or 3000 image analyser were analysed using Fuji Image Gauge software (Fuji Film Co. Ltd). For each protein these area values were added together to give a ‘total gradient' or ‘total DRM' protein value. The protein band intensities for the S1 and P1M samples were calculated by taking into account the volume that was loaded onto the gel compared with the original S1 and P1M sample size. Amounts of ¹²⁵I-NGF in each fraction were determined using a gamma counter. For each protein (or ¹²⁵I-NGF), the amount in the floating DRM peak was calculated and expressed as a percentage of that particular protein in all cell fractions. P values were calculated using the students t-test.