Inhibition of sphingolipid synthesis affects kinetics but not fidelity of L1/NgCAM transport along direct but not transcytotic axonal pathways.

Abstract

Glycosphingolipids are constituents of lipid rafts which might function in sorting apical and axonal cargoes in the trans-Golgi network. In fact, two GPI-linked proteins, Thy1 and PrPC, require lipid raft lipids for sorting to the axon. It was previously shown that inhibition of glycosphingolipid synthesis by FumonisinB1 (FB1) impairs axon outgrowth but not axon specification, leading to the hypothesis that formation of axonally-targeted vesicles is coupled to sphingolipid synthesis. Since the axonal cell adhesion molecule L1/NgCAM can partition into membrane rafts biochemically, we asked whether correct targeting to the axon is FB1-sensitive, similarly to GPI-linked proteins. We previously showed that cultured hippocampal neurons use more than one trafficking pathway to the axon: a transcytotic pathway and a direct pathway. We show here that reducing raft lipid levels does not disrupt axonal targeting of L1/NgCAM along either pathway. Unexpectedly, FB1 selectively slowed the kinetics of surface expression of a truncated NgCAM using the direct pathway, but not of NgCAM using the transcytotic pathway. Therefore, the formation and/or transport of a subset of axonally-targeted vesicles are coupled to sphingolipid synthesis. Our results yield a mechanism for the axon outgrowth defect observed in FB1.

Full text

Page 1

Inhibition of sphingolipid synthesis affects kinetics but not
fidelity of L1/NgCAM transport along direct but not
transcytotic axonal pathways
Michael C. Chang,1Dolora Wisco,1Helge Ewers, Caren Norden, and Bettina Winckler*
Department of Neuroscience, University of Virginia, 409 Lane Road, MR4-6112, Charlottesville, VA 22908, USA
Received 13 May 2005; revised 28 September 2005; accepted 11 November 2005
Available online 20 December 2005
Glycosphingolipids are constituents of lipid rafts which might function
in sorting apical and axonal cargoes in the trans-Golgi network. In fact,
two GPI-linked proteins, Thy1 and PrPC, require lipid raft lipids for
sorting to the axon. It was previously shown that inhibition of
glycosphingolipid synthesis by FumonisinB1 (FB1) impairs axon
outgrowth but not axon specification, leading to the hypothesis that
formation of axonally-targeted vesicles is coupled to sphingolipid
synthesis. Since the axonal cell adhesion molecule L1/NgCAM can
partition into membrane rafts biochemically, we asked whether correct
targeting to the axon is FB1-sensitive, similarly to GPI-linked proteins.
We previously showed that cultured hippocampal neurons use more
than one trafficking pathway to the axon: a transcytotic pathway and a
direct pathway. We show here that reducing raft lipid levels does not
disrupt axonal targeting of L1/NgCAM along either pathway. Unex-
pectedly, FB1 selectively slowed the kinetics of surface expression of a
truncated NgCAM using the direct pathway, but not of NgCAM using
the transcytotic pathway. Therefore, the formation and/or transport of
a subset of axonally-targeted vesicles are coupled to sphingolipid
synthesis. Our results yield a mechanism for the axon outgrowth defect
observed in FB1.
D 2005 Elsevier Inc. All rights reserved.
Keywords: L1/NgCAM; Lipid raft; Axonal targeting; Transcytosis;
Fumonisin B1
Introduction
Cellular membranes are made up of not only a large variety of
membrane-associated proteins, but also of many types of lipids.
The role of lipids in cellular processes is receiving increasing
attention, not last because of a large variety of metabolic disorders
in lipid metabolism. For instance, a number of inherited metabolic
disorders result from defects in the lysosomal enzymes involved in
the degradation of ceramide-derived glycosphingolipids (GSLs)
(Ginzburg et al., 2004). GSLs are found at the highest levels in the
brain, particularly in neurons (Ginzburg et al., 2004) and therefore
much attention has been focused on understanding the role of
GSLs in neuronal function. A number of years ago, Futerman and
co-workers showed that inhibition of ceramide synthesis with
Fumonisin B1 (FB1) in cultured hippocampal neurons leads to
shorter axons and less axonal branching (Harel and Futerman,
1993; Schwarz et al., 1995). They proposed that the formation and/
or transport of axonally targeted vesicles are coupled to sphingo-
lipid synthesis.
How could GSLs function in axonal membrane transport?
GSLs, sphingomyelin, and cholesterol are constituents of lipid
rafts, microdomains present in many cellular membranes (Simons
and Ikonen, 1997). Even though the exact size and nature of lipid
rafts are controversial (Munro, 2003), much evidence has
accumulated over the last decade to point at an important role
for lipid rafts in many biological processes, including signaling and
polarized sorting (Fullekrug and Simons, 2004). In epithelial cells,
several apical proteins are missorted when rafts are disrupted
pharmacologically using either inhibitors of cholesterol synthesis
or inhibitors of ceramide synthesis, such as Fumonisin B1
(reviewed in Helms and Zurzolo, 2004; Holthuis et al., 2003;
Ikonen, 2001). Some apical proteins (including the viral hemag-
glutinin and many GPI-linked proteins) associate preferentially
with raft lipids using biochemical assays (Lisanti and Rodriguez-
Boulan, 1990). These observations have led to a model in which
apical proteins preferentially associate with rafts in the Golgi and
trans-Golgi network (TGN), leading to cargo enrichment. It was
recently shown that oligomerization of GPI-linked proteins aug-
ments raft association as well as faithful apical targeting in MDCK
cells (Paladino et al., 2004). Additional enrichment, therefore, can
take place through oligomerization of raft components by putative
clustering proteins. This cross-linking clusters small rafts into
bigger rafts and increases the affinity of weakly raft-associating
apical cargo for raft domains (Rajendran and Simons, 2005;
Schuck and Simons, 2004). For several apically-targeted proteins,
glycosylation sites in the lumenal domain are required for correct
1044-7431/$ - see front matter D 2005 Elsevier Inc. All rights reserved.
doi:10.1016/j.mcn.2005.11.006
* Corresponding author.
E-mail address: BWinckler@virginia.edu (B. Winckler).
1These authors contributed equally to this work.
Available online on ScienceDirect (www.sciencedirect.com).
www.elsevier.com/locate/ymcne
Mol. Cell. Neurosci. 31 (2006) 525 – 538

Page 2

targeting. This observation has led to the idea that lumenal lectins
might act as raft-associated clustering agents and cross-link raft-
associated proteins into clustered raft domains (Rajendran and
Simons, 2005; Schuck and Simons, 2004). These apical cargo-rich
domains are thought to subsequently bud off and give rise to
apically-targeted vesicles. The molecular machinery responsible
for budding of raft domains is not yet understood (Rodriguez-
Boulan et al., 2005), but lipid rafts might be instrumental in the
budding reaction as well (Rajendran and Simons, 2005; Schuck
and Simons, 2004).
In neurons, even less is known but similar mechanisms might
also be at play. Inhibition of ceramide synthesis with Fumonisin B1
(FB1) leads to mistargeting of the GPI-linked protein Thy1
(Ledesma et al., 1999; Ledesma et al., 1998) and most recently
PrPC (Galvan et al., 2005), suggesting a role for ceramide-derived
lipids in axonal targeting. On the other hand, no mistargeting after
FB1 treatment was observed in another study for three different
kinds of proteins: the axonal GAP43 which associates with the
cytoplasmic leaflet of the plasma membrane via palmitoylation, the
synaptic vesicle protein synaptophysin, and the somatodendritic
cytoskeletal protein MAP2 (Schwarz et al., 1995). None of the
proteins studied so far have been transmembrane proteins localized
to the axonal plasma membrane. We therefore asked whether an
axonal transmembrane protein that can associate with rafts
biochemically is FB1-sensitive for axonal targeting.
The plasma membrane of many cells is compartmentalized with
different membrane proteins occupying different sites (Mellman,
1995). In neurons, axons and dendrites display different membrane
proteins on their surfaces. How this differential distribution of
membrane proteins to axons and dendrites is achieved is under
active investigation by many labs (Craig and Banker, 1994; Horton
and Ehlers, 2003; Winckler, 2004; Winckler and Mellman, 1999).
Our lab is studying the pathways underlying the axonal accumu-
lation of the cell adhesion molecule L1/NgCAM. We found
previously that NgCAM can travel to the axon by an indirect
transcytotic pathway via the somatodendritic plasma membrane
(Wisco et al., 2003) followed by endocytosis and sorting to the
axon from the endosome. Sequences in the cytoplasmic tail are
required for trafficking along the transcytotic pathway. Truncation
of the cytoplasmic tail of NgCAM (NgCAM(CT3)) does not
impair axonal accumulation, but changes the pathway traveled:
NgCAM(CT3) reaches the axonal surface by a direct pathway
(Wisco et al., 2003). Endogenous axonal cargoes using the direct
pathway have not yet been identified, but presumably exist.
Therefore, two distinct pathways to the axon exist. Transcytotic
proteins (such as NgCAM) are sorted into somatodendritically-
directed vesicles because they contain appropriate cytoplasmic tail
signals which appear to be recognized by cytoplasmic adaptor
proteins (Anderson et al., 2005). NgCAM(CT3), traveling on the
direct pathway, lacks cytoplasmic tail signals and depends on
lumenal signals for axonal sorting (Sampo et al., 2003). Since
NgCAM(CT3) cannot by itself interact with cytoplasmic adaptors
to achieve inclusion into axonally-directed vesicles, we asked
whether – similarly to GPI-linked proteins in MDCK cells and
neurons – association with lipid rafts might be important for
axonal accumulation of NgCAM(CT3).
Fig. 1. Low DRM association of NgCAM. (A) Cultures were infected with AdNgCAM and processed for DRM analysis as described in Experimental methods.
Fractions were collected from the gradient from the top. The same blot was probed sequentially for caveolin, transferrin receptor, and NgCAM. The position of
DRM-associated material (fractions 5–7) was defined by the floating populations of the raft markers caveolin and GM1 (cholera toxin B-binding material). The
endogenous caveolin signal is faint, but detectable in fractions 5–7/8 (arrows) as well as in the starting material fractions 16,17. TfR does not associate with
DRMs and is not detected in fractions 5–7. NgCAM is also not detected in fractions 5–7. (B) Fractions were combined as indicated above the lanes. On long
overexposures, a small proportion (<5%) of NgCAM is reproducibly associated with DRM fractions 5–8. We note that all known proteolytic forms of NgCAM
(200 kDa, 140 kDa, and 80 kDa) associate with DRMs to similar degrees.
M.C. Chang et al. / Mol. Cell. Neurosci. 31 (2006) 525–538
526

Page 3

Results
Detergent resistant membrane association of L1/NgCAM
Since L1 associates with rafts biochemically (Nakai and
Kamiguchi, 2002) in detergent resistant membrane (DRM)
fractions, we asked if axonal targeting of L1/NgCAM is dependent
on glycosphingolipid synthesis, as shown for Thy1 (Ledesma et al.,
1998) and PrPC(Galvan et al., 2005). We first wanted to test if L1/
NgCAM has the potential to associate with DRMs in our cultured
hippocampal neurons. We therefore fractionated neuronal cultures
expressing NgCAM on sucrose gradients after detergent extraction
(see Experimental methods). The lipid raft markers caveolin and
GM1 were used as positive markers to identify DRM fractions
while the non-raft protein transferrin receptor was used as a
negative control. Under stringent conditions (see Experimental
methods), non-raft proteins (like transferrin receptor) are com-
pletely depleted from DRM fractions while potentially raft-
associating proteins (like caveolin) are partially or completely
enriched in DRM fractions. The exact proportion of DRM
enrichment depends on both the intrinsic capacity of a protein to
associate with raft lipids and on the stringency of the DRM
preparation procedure. We chose high stringency conditions under
which non-raft proteins cannot be detected in the DRM fractions,
even when overexpressed. Any NgCAM detected in the DRM
fractions is therefore an indication that NgCAM can biophysically
interact with raft lipids in vitro. The same blot was sequentially
probed for NgCAM, caveolin, and transferrin receptor. Cholera
toxin B subunit (CT-B) specifically binds GM1 and is commonly
used for GM1 detection (Lamaze et al., 2001; Nakai and
Kamiguchi, 2002). GM1 was detected on dot blots using HRP-
cholera toxin B. Transferrin receptor was found only in the bottom
fractions 15–17 in our gradient system (Fig. 1A), consistent with
non-raft association. Caveolin and GM1, on the other hand, can be
detected additionally in the low density fractions 5–7/8 (Fig. 1A,
arrows). Based on the fractionation of GM1 and caveolin, we
designated fractions 5–7/8 as DRM. We could not readily detect
NgCAM in DRM fractions under these conditions. We therefore
pooled fractions 1–4, 5–8, 9–12, and 13–16 into four lanes. On
long exposures of NgCAM blots, we now reproducibly detected
NgCAM in the DRM fraction (Fig. 1B). This is in agreement with
previous observations (Nakai and Kamiguchi, 2002). In this
previous study, less stringent detergent conditions were used and
a larger proportion of L1/NgCAM fractionated with DRMs in a
developmentally regulated fashion. Under our stringent condition
for DRM association, about 50% of caveolin and GM1 float to the
DRM fraction. NgCAM, on the other hand, is associated with
DRM fractions to a lower degree (<5%). As shown by others, the
Fig. 2. FB1 does not disturb axonal localization of L1. Control cultures (A, B) or cultures treated with FB1 (C, D) were fixed on day 9 after plating and stained
with an antibody against L1 (A, C) and FITC-Cholera toxin B (B, D) without permeabilization. Arrows designate axons; arrowheads designate dendrites.
M.C. Chang et al. / Mol. Cell. Neurosci. 31 (2006) 525–538
527

Page 4

extent of DRM association of the same protein is highly dependent
on the exact experimental procedure used (Brown, 2001; London
and Brown, 2000; Munro, 2003; Shogomori and Brown, 2003).
We chose stringent conditions so that we could rule out the
possibility that the presence of NgCAM in low-density fractions
was due to incomplete solubilization of plasma membranes. Our
ability to detect NgCAM in DRM fractions, therefore, shows that it
has the capacity to fractionate into a lipid raft environment.
Transferrin receptor, on the other hand, is always undetectable in
the DRM fractions (see also below).
Axonal accumulation of L1 is not disrupted by Fumonisin B1
We asked next if the axonal targeting of endogenous L1 was
disrupted by decreasing lipid raft lipids with Fumonisin B1 (FB1).
FB1 inhibits ceramide synthase and thereby affects the synthesis of
sphingomyelin and glycosphingolipids which are lipid raft
constituents. FB1 has been used successfully on cultured hippo-
campal neurons (Harel and Futerman, 1993; Ledesma et al., 1998;
Schwarz et al., 1995) to deplete ceramide-derived lipids. FB1-
treated neurons still elaborate axons and dendrites, but strikingly
axonal growth and branching are decreased (Harel and Futerman,
1993; Schwarz et al., 1995). Dotti and co-workers also showed that
the synthesis of sphingomyelin, one of the ceramide-derived raft
lipids, is crucial for axonal sorting of Thy1 (Ledesma et al., 1999).
In contrast, chronic depletion of cholesterol, another lipid raft
constituent, is highly toxic to neurons and only modest decreases in
cholesterol levels (up to 50%) can be achieved before toxicity
becomes significant (El-Husseini et al., 2001). If cholesterol is
depleted for short times using methyl-h-cyclodextrin, endosomal
trafficking of cholera toxin is impaired (Shogomori and Futerman,
2001). To test whether axonal sorting of the endogenously-
Fig. 3. FB1 leads to >80% reduction of GM1 levels. (A) Neuronal cultures treated with increasing amounts of FB1 (as indicated along the bottom) were fixed
and stained with FITC-cholera toxin (top panels). Staining intensity decreased progressively with higher FB1 doses. Phase images are shown in the bottom
panels. We note that we detect cholera toxin staining equally on axons and dendrites in agreement with Shogomori et al. (1999). (B, C) Levels of the
ganglioside GM1 were quantified by either determining average pixel intensity along axons stained with FITC-Cholera toxin B (B) or dot blots using HRP-
Cholera toxin B of lysates prepared from control or cultures treated with 25 AM FB1 (C). In panel B, random microscope fields were photographed with
identical exposures. The average pixel intensity along axons in the field was determined by IPLab software. The horizontal line shows the average. In panel C,
equal protein concentrations of control or FB1 lysates were spotted onto nitrocellulose at the dilutions indicated. GM1 was detected using HRP-cholera toxin B.
M.C. Chang et al. / Mol. Cell. Neurosci. 31 (2006) 525–538
528

Page 5

expressed L1 was sensitive to FB1, we treated cultured hippo-
campal neurons with FB1, essentially as described before
(Ledesma et al., 1998). The appearance of the Golgi was not
grossly disturbed by FB1-treatment at the light microscope level
(Supplementary Fig. 1). In untreated control cultures, L1 is highly
enriched along axons but largely undetectable on somata and
dendrites (Fig. 2A; arrows designate axons; arrowheads dendrites).
After treatment with FB1, L1 is still found highly enriched along
axons (Fig. 2C) and cells where L1 was present on dendrites as
well as axons were never found. Therefore, no missorting was
observed.
FB1 has been used successfully on cultured hippocampal
neurons to deplete ceramide-derived lipids, including sphingomye-
lin and glycosphingolipids by 75–80% (Harel and Futerman,
1993; Ledesma et al., 1998). To determine the decrease in
ceramide-derived lipid under our conditions, we determined the
levels of the ceramide-derived ganglioside GM1. Control cultures
stained brightly on all processes with FITC-CT-B (Fig. 2B) while
FB1-treated cultures showed significantly reduced staining (Fig.
2D) in a dose-dependent manner (Fig. 3A). The FITC-CT-B
surface levels were quantified using IPLab software on identically-
processed images from FB1-treated (25 AM) or untreated cultures.
The average decrease of FITC-CT-B staining after FB1 treatment is
80% (Fig. 3B). We additionally quantified GM1 levels on dot blots
from cell lysates prepared from control cultures or FB1-treated
cultures. Serial dilutions of each of the lysates were spotted onto a
filter after normalizing to protein content and probed with HRP-
CT-B. FB1 treatment resulted in a >80% decrease of GM1 levels
(Fig. 3C). We conclude that L1 sorting to the axon is insensitive to
an 80% decrease of a ceramide-derived lipid species.
The polarized localization of LDLR, NgCAM, and NgCAM(CT3)
are not disrupted by FB1
Endogenous expression of L1 starts shortly after axon initiation
(Peretti et al., 2000) when FB1 treatment is first begun. We tested if
Fig. 4. FB1 does not disturb the polarized surface distribution of LDLR, NgCAM, or NgCAM(CT3). Control cultures (A, C, E) or FB1-treated cultures (B, D, F)
were infected with AdLDLR (A, B), AdNgCAM (C, D), or AdNgCAM(CT3) (E, F) and fixed 24 to 36 h later. The corresponding phase images are shown as
insets. Arrows designate axons, arrowheads dendrites. Faint soma staining is due to anti-NgCAM antibody background (see also Jareb and Banker, 1998).
Dendrites are not stained above background.
M.C. Chang et al. / Mol. Cell. Neurosci. 31 (2006) 525–538
529

Page 6

axonal targeting is sensitive to FB1 when axonal proteins are
expressed acutely by adenoviral vectors at days 7–8 in vitro (DIV),
i.e. after FB1 treatment is complete. To this end, we infected
neurons at 7 DIV with a defective adenovirus expressing NgCAM,
the chick homolog of L1. NgCAM accumulates preferentially on
the axon after viral expression (Jareb and Banker, 1998; Vogt et al.,
1996; Winckler et al., 1999). The virally-expressed NgCAM
protein can then be specifically detected with a species-specific
antibody (Vogt et al., 1996). As a control, cells were infected with
an adenovirus expressing the somatodendritic protein LDLR (Fig.
4A). Treatment with FB1 did not affect the steady-state distribution
of either LDLR (Fig. 4B; compare to A) or NgCAM (Fig. 4D;
compare to C). For both markers, over 80% of cells showed correct
polarized distribution with or without FB1.
Fig. 5. FB1 delays axonal surface expression of NgCAM(CT3). (A) Decreased surface expression of NgCAM(CT3) in FB1-treated cells. Surface expression of
NgCAM(CT3) was quantified in control (n) and FB1-treated (r) cells 32 h after infection with AdNgCAM(CT3) by determining the average pixel intensity
along axons using IPLab software after identical processing of randomly photographed cells. Please note that only cells with detectable surface expression were
included. Each data point is the average axon pixel intensity of one cell. The black line demarks the average. (B–D) FB1 delays surface delivery of
NgCAM(CT3). (B) Cells were infected with AdNgCAM and fixed between 18 and 27 h after infection. Surface and intracellular pools were detected with
different fluorophores (see Supplementary Fig. 2). At 22 h postinfection (a, b), NgCAM is not detectable on the surface (a) even though bright intracellular
staining is detectable (b). At 24 h postinfection (c, d), NgCAM is detectable both intracellularly (d) as well as on the axonal cell surface (c). Arrows designate
axons. Please note that intracellular staining is detectable in dendrites and axons while surface staining is restricted to the axon. (C) Control cultures (filled
symbols) or FB1-treated cultures (open symbols) were infected with AdLDLR (.), AdNgCAM (r), or AdNgCAM(CT3) (n) and fixed 18 to 32 h later as
indicated on the x-axis. Surface and intracellular pools of the expressed proteins were detected with different fluorophores (as in panel B). Cells showing
intracellular staining were scored for surface expression. One representative experiment is shown. (D) Quantification of kinetic delay of NgCAM(CT3). The
percentage of cells with detectable surface expression of NgCAM (at 24 h), LDLR (at 22 h), or NgCAM(CT3) (at 32 h) were counted in FB1-treated cultures
and normalized to control cultures. The average of three (LDLR) or four (NgCAM and NgCAM(CT3)) independent experiments is shown. Bar demarks the
standard error of the mean.
M.C. Chang et al. / Mol. Cell. Neurosci. 31 (2006) 525–538
530

Page 7

We recently showed that cultured hippocampal neurons use
both indirect (transcytotic) as well as direct pathways for sorting
proteins to the axon (Wisco et al., 2003). Kinetic analysis of
NgCAM axonal targeting suggested that NgCAM does not travel
directly from the trans-Golgi network (TGN) to the axon, but rather
is first transported to the somatodendritic domain, and then
endocytosed and transported to the axon from the endosome
(Wisco et al., 2003). This is reminiscent of transcytotic targeting of
apical proteins in hepatocytes (Tuma and Hubbard, 2003) and of at
least some apical proteins in MDCK cells (Polishchuk et al., 2004).
Whenthecytoplasmictail ofNgCAMwasdeleted(NgCAM(CT3)),
axonal targeting was preserved (Sampo et al., 2003; Wisco et al.,
2003). However – unlike full length NgCAM – NgCAM(CT3)
traveled on a direct pathway to the axon (Wisco et al., 2003) and did
not transiently appear on the somatodendritic domain. We can
therefore use NgCAM(CT3) to probe specifically the direct axonal
pathway. Since NgCAM(CT3) lacks cytoplasmic sorting signals
and travels directly from the TGN to the axon, we hypothesized that
it might depend on lipid rafts and might be sensitive to FB1
treatment, even if NgCAM is insensitive. To test this idea, we
infected FB1-treated and control cells with adenovirus expressing
NgCAM(CT3) and stained the surface with an anti-NgCAM
antibodyafterfixation.SimilarlytoL1andNgCAM,NgCAM(CT3)
accumulates preferentially on the axon after FB1 (>90% of cells),
similarly to controls (Fig. 4F; compare to E). Therefore, unlike
Thy1, neither NgCAM nor NgCAM(CT3) is missorted after FB1
treatment.
FB1 slows the kinetics of surface expression of NgCAM(CT3)
During the previous experiments, we noticed that the surface
expression of NgCAM(CT3) appeared fainter after FB1 than in
controls, even though it is still axonally enriched. In order to
investigate whether less NgCAM(CT3) accumulated on the surface
after FB1, we photographed randomly chosen cells with detectable
NgCAM(CT3) on the surface and used IPLab software to quantify
the average pixel intensity along axons. The average pixel intensity
along axons was twofold lower in FB1-treated cells (r) compared
to control cells (n) (Fig. 5A). This observation raised the
possibility that FB1 slowed the kinetics of surface delivery of
NgCAM(CT3) without affecting the fidelity of axonal sorting.
In order to test this hypothesis, we infected cultures with
adenoviruses encoding NgCAM, NgCAM(CT3), or LDLR and
fixed cells at different times to follow surface expression kinetically.
We used a sandwich staining technique to identify cells which
express detectable amounts of the test protein intracellularly with
one fluorophore and then score whether or not the same cell
expressed the test protein on the cell surface with a second
fluorophore (Supplementary Fig. 2, see Experimental methods).
Few cells with intracellular expression of NgCAM were detectable
shorter than 18 h postinfection. In cells where intracellular
expression was detectable at 18 to 20 h, surface expression was
rarely detectable (Figs. 5B, a, b). By 27 h postinfection, >90% of
intracellularly-expressing cells showed surface expression of
NgCAM along axons (Figs. 5B, c, d and C, r).
The same kinetic analysis was performed on cells with or
without FB1 treatment. NgCAM appeared on the axonal cell
surface with the same kinetics regardless of treatment (Fig. 5C, q
vs. r). Similarly, the kinetics of LDLR transport to the
somatodendritic surface was not affected by FB1 (Fig. 5C, o vs.
.). The first detectable intracellular expression of NgCAM(CT3)
occurred more slowly than that of NgCAM and LDLR, even in
control cells. Few, if any cells had detectable intracellular staining
even at 27 h postinfection with AdNgCAM(CT3) (Fig. 5C, n). At
32 h postinfection, though, >85% of intracellularly-expressing cells
also had detectable cell surface expression of NgCAM(CT3) (Fig.
5C, n), indicating similar surface transport kinetics once intracel-
lular staining was detectable. In the FB1-treated cells, on the other
hand, surface expression of NgCAM(CT3) was greatly delayed. At
32 h postinfection, only 30% of intracellularly-expressing neurons
showed surface expression of NgCAM(CT3) compared to 85% in
the control cells (Fig. 5C, g vs. n). To quantify this effect across
multiple experiments, the percentage of surface-expressing cells
after FB1 was normalized to control cells for NgCAM, LDLR, and
NgCAM(CT3) for four separate experiments (Fig. 5D). The
average normalized ratio for NgCAM and LDLR was ¨1
indicating no kinetic differences of surface expression (Fig. 5D).
On the other hand, the average normalized ratio for NgCAM(CT3)
was 0.58, indicating kinetic delay of NgCAM(CT3) after FB1
treatment.
One possible explanation for decreased surface detection of
NgCAM(CT3) is that less total NgCAM(CT3) accumulated in
FB1-treated cultures because of slowed synthesis or increased
turnover. Therefore, we performed Western blot analysis of lysates
Fig. 6. Expression level and DRM association of NgCAM(CT3). (A)
Control cultures (right lane) and FB1-treated cultures (left lane) were
infected with AdNgCAM(CT3) for 34 h. Lysates were prepared and equal
amounts of protein loaded per lane. Full length NgCAM(CT3) is about 200
kDa. Proteolytically-processed forms of 140/135 kDa are also detected at
comparable levels. (B) Cultures were infected with AdNgCAM(CT3) and
processed for DRM analysis as described in Experimental methods.
Fractions were collected from the gradient from the top. The position of
DRM-associated material (fractions 2–3) was defined by the floating
populations of the raft markers caveolin and GM1 (cholera toxin B-binding
material).
M.C. Chang et al. / Mol. Cell. Neurosci. 31 (2006) 525–538
531

Page 8

prepared from NgCAM(CT3)-infected cultures with or without
FB1. Equal amounts of total protein were loaded. No differences in
Western signal against NgCAM(CT3) could be detected (Fig. 6A).
Therefore, depletion of ceramide-derived lipid species with FB1
leads to a kinetic delay of transport to the cell surface for the
directly-transported NgCAM(CT3), but not for the transcytosing
NgCAM or the somatodendritic LDLR.
NgCAM(CT3) still associates with DRMs
Since NgCAM and NgCAM(CT3) showed different kinetic
sensitivity to FB1, we tested if NgCAM(CT3) still partitioned into
DRMs to the same degree as NgCAM. Neurons were infected with
NgCAM(CT3) adenoviruses and Triton X-100 lysates fractionated
on sucrose gradients, as described above. As before, the lipid raft
markerscaveolinandGM1wereusedaspositivemarkerstoidentify
DRM fractions. Fractions were collected from the top in the
following manner: Lane 1 corresponds to fractions 1–4, lane 2 to
fractions 5/6, lane 3 to fractions 7/8, lane 4 to fractions 9–11, lane 5
to fractions 12–14, lane 6 to fraction 15, and lane 7 to fraction 16.
We found that NgCAM(CT3) was reproducibly detectable in DRM
fractions (lane 2/3) (Fig. 6B). Transferrin receptor was overex-
pressed using adenovirus (AdTfR) to control for the possibility that
highoverexpressioncouldresultindetectabilityofanon-raftprotein
in the DRM fractions. No TfR was detected in the DRM fractions
even on long exposures of the blot. Therefore, no change in DRM
fractionation behavior could be observed.
Slowed NgCAM(CT3) resides in Golgi/TGN compartments
Next, we asked where the slowed NgCAM(CT3) was found in
FB1-treated cells. We hypothesized that the slow step might occur
in the Golgi/TGN, rather than in the ER, since lipid rafts lipids are
largely synthesized in post-ER compartments in mammalian cells
(Futerman and Riezman, 2005). We therefore co-stained FB1-
treated cells expressing NgCAM(CT3) with markers for Golgi and
TGN and imaged the cells by laser-scanning confocal microscopy
(Fig. 7). Single confocal sections are shown. We found that the
intracellular pool of NgCAM(CT3) overlapped greatly with
GRASP65 (a Golgi marker; Figs. 7A and B; two examples are
shown) and TGN38 (a TGN marker; Figs. 7C and D; two examples
of each are shown). Some regions of non-overlap were also
observed. These observations are consistent with the notion that
FB1 slows the transit of NgCAM(CT3) through the Golgi/TGN
compartments. Subsequent steps (i.e. vesicle formation, transport,
and fusion) might additionally be affected.
Fig. 7. Co-localization of NgCAM(CT3) with Golgi and TGN in FB1-treated cells. (A, B) FB1-treated cells expressing NgCAM(CT3) were double-stained for
NgCAM(CT3) (green) and the Golgi marker GRASP65 (red). Merged images are shown in the left panel. (C, D) FB1-treated cells expressing NgCAM(CT3)
were double-stained for NgCAM(CT3) (green) and the TGN marker TGN38 (red). Merged images are shown in the left panel. Cells were imaged by laser
scanning confocal microscopy.
M.C. Chang et al. / Mol. Cell. Neurosci. 31 (2006) 525–538
532

Page 9

Intracellular NgCAM(CT3) shows increased Triton-insolubility
We hypothesized that the FB1-sensitivity of NgCAM(CT3)
transport might arise becauseNgCAM(CT3) needs toassociate with
glycosphingolipid-containing lipid rafts in the Golgi/TGN for
efficient packaging and delivery to the plasma membrane. Previous
work has shown that raft-associated proteins are Triton X-100
insoluble at 4-C, but not at 37-C (Brown and Rose, 1992). We
therefore tested if NgCAM(CT3) was Triton-insoluble in the Golgi/
TGN at 4-C. 28–30 h after infection with AdNgCAM(CT3), we
extracted live cells in 0.5% Triton-X 100 either on ice or at 37-C for
3 min, fixed the extracted cells, and stained with an antibody against
NgCAM. In non-extracted cells, NgCAM(CT3) was observed on
axons as well as in the Golgi region of the soma (Fig. 8A). In the
Fig. 8. Differential Triton-solubility of NgCAM(CT3) in the soma and axon. (A) NgCAM(CT3) in a non-extracted cell can be found in the Golgi region of the
soma as well as strongly along the axon. The cell was permeabilized after fixation. (B, C) NgCAM(CT3) staining can be observed in the Golgi region and
faintly on axons after live Triton X-100-extraction at 4-C (B), but not after extraction at 37-C (C). (D) Control cells (grey bars) or cells extracted live in 0.5%
Triton X-100 (TX) (black bars) were photographed and processed identically. In 20–30 cells per experiment, the average pixel intensity of NgCAM(CT3)
staining was determined in the Golgi region (left bars) and along the axon (right bars). The average staining intensity in TX-extracted cells was normalized to
the intensity in unextracted control cells. Three independent experiments were quantified. Three examples of the staining intensities observed are shown under
the corresponding bars.
M.C. Chang et al. / Mol. Cell. Neurosci. 31 (2006) 525–538
533

Page 10

extractedcells,stainingwasgreatlyreduced.Nevertheless,weeasily
observed Triton-insoluble populations of NgCAM(CT3). The most
common staining pattern of extracted cells was perinuclear soma
staining, reminiscent of Golgi/TGN (Fig. 8B). In order to test
whether the Triton-insoluble staining in the soma was associated
with Golgi/TGN, we double-stained with three different markers for
Golgi and TGN. We found that the Golgi markers themselves were
not resistant to Triton extraction and no staining remained (data not
shown). We also observed faint Triton-insoluble NgCAM(CT3)
staininginapunctatepatternalongaxons.Extractionat37-Clargely
eliminated NgCAM(CT3) staining (Fig. 8C) and only very occa-
sional faint staining was observed.
In order to quantify the extent of Triton-insolubility of
NgCAM(CT3), cells were photographed with identical exposure
settings and the average intensity of NgCAM(CT3) staining
determined along axons and in the Golgi region of extracted and
non-extracted controls. The average intensities of randomly chosen
cells are plotted in a bar graph in Fig. 8D and three examples
showing the observed range of intensities are shown below each
category. The average intensity of NgCAM(CT3) remaining in the
Golgi region after Triton-extraction was 50% of unextracted
controls. For the axonal staining, on the other hand, only 7% of
the control intensity remained after extraction (Fig. 8D). These
observations show that a large proportion of NgCAM(CT3) is
Triton-insoluble in the Golgi region, suggesting that it might
partition into Triton-insoluble lipid rafts in the Golgi/TGN.
Interestingly, the axonal pool of NgCAM(CT3) was overwhelm-
ingly Triton-soluble. Therefore, the insolubility of NgCAM(CT3)
occurs transiently on the biosynthetic pathway and does not persist
on the plasma membrane.
Discussion
The role of lipid rafts in cell functioning has been under intense
investigation in many cell types. Many examples exist for the role
of lipid rafts as signaling platforms, especially in lymphocytes
(Harder, 2004) but also in neurons (Guirland et al., 2004). But
evidence has also accumulated that rafts can serve as lateral sorting
platforms (probably in the Golgi/TGN) for apically-directed cargo
in epithelial cells (Fullekrug and Simons, 2004; Helms and
Zurzolo, 2004). Much of this work has relied on depleting
components of lipid rafts, either cholesterol or ceramide which is
the precursor for both glycosphingolipids as well as for sphingo-
myelin. For instance, inhibition of ceramide synthase using
Fumonisin B1 leads to mistargeting of some apical proteins,
especially GPI-linked proteins, in Fisher Rat Thyroid (FRT) cells
(Lipardi et al., 2000) and MDCK cells (Mays et al., 1995).
Cholesterol depletion similarly leads to missorting of some apical
proteins (HA in MDCK cells (Keller and Simons, 1998);
aminopeptidase in enterocytes (Hansen et al., 2000)). Much
interest is now focused on identifying raft-associated clustering
agents that can increase the affinity of weakly raft-associating
cargo molecules for raft domains. Candidates for such clustering
agents are MAL and some members of the annexin family
(reviewed in Rajendran and Simons, 2005; Schuck and Simons,
2004). On the other hand, there is also raft-independent sorting to
the apical domain (Jacob and Naim, 2001; Nelson and Yeaman,
2001) as well as basolateral sorting of some raft-associated proteins
(Sarnataro et al., 2002). These studies together suggest that
different pathways can be differentially dependent on lipid rafts,
that the same pathways in different cell types can be differentially
dependent on lipid rafts, and that lipid rafts, even if necessary for
some cargoes, are not sufficient for apical targeting, but require
other sorting machinery.
In neurons, the picture is even less clear-cut, at least in part
because only very few studies have investigated the role of rafts in
polarized sorting. Missorting was reported for the GPI-linked
proteins Thy1 and PrPCafter FB1 treatment in cultured hippo-
campal neurons (Ledesma et al., 1998; Galvan et al., 2005). Three
different kinds of proteins showed no missorting after FB1
treatment: the axonal GAP43 which associates with the cytoplas-
mic leaflet of the plasma membrane via palmitoylation, the
synaptic vesicle protein synaptophysin, and the somatodendritic
cytoskeletal protein MAP2 (Schwarz et al., 1995). None of the
proteins studied until now had been axonal plasma membrane
transmembrane proteins. We therefore asked whether L1/NgCAM,
an axonal transmembrane protein which can associate with DRMs
(Nakai and Kamiguchi, 2002), is FB1-sensitive for axonal
targeting. Additionally, we probed the FB1-sensitivity of two
distinct axonal pathways, a transcytotic pathway used by NgCAM
and a direct pathway used by NgCAM(CT3). Our results show no
missorting along either pathway.
Another study also did not observe any missorting of polarized
proteins (Harel and Futerman, 1993; Schwarz et al., 1995). The
authors found that FB1 treatment caused a substantial reduction in
axonal growth and axonal branching (Harel and Futerman, 1993;
Schwarz et al., 1995). In a follow-up study, the same group showed
that ceramide had to be metabolized to glucosylceramide in order
to sustain axonal growth (Schwarz and Futerman, 1997), and that
ceramide was not able to substitute. Since ceramide is involved in
multiple signaling pathways whose disruption could potentially
impair axon outgrowth, this result indicates that disruption of
ceramide-based signaling by FB1 does not account for the axon
growth defect. Rather than signaling by ceramide per se,
glucosylceramide and/or its metabolites (i.e. gangliosides) are
required for maximal axonal growth. In fact, these lipids might be
specifically required for one (or more) crucial intracellular
transport pathway(s) that supplies membrane to the axon. The
data presented in this work provide evidence in support of this
hypothesis. Using a kinetic approach, we find that FB1 treatment
leads to the selective slowing of a particular intracellular pathway.
The direct axonal pathway (used by a truncated NgCAM) is slowed
down, while the somatodendritic pathway (used by LDLR) or the
transcytotic pathway (used by NgCAM) is not affected. Our kinetic
studies therefore show that the transport efficiencies of different
pathways are in fact differentially dependent on sphingolipid
synthesis in cultured hippocampal neurons. Since the effect of FB1
is kinetic, previous studies using analysis of steady-state distribu-
tion did not report disturbances of several axonal proteins. Our
results suggest a mechanism for the observed axonal growth defect
in FB1-treated neurons, observed by Futerman and colleagues
(Schwarz et al., 1995). Since axonal proteins are not missorted
along either the transcytotic or direct pathways, axons are still
correctly specified and elaborated, i.e. neuronal polarity is not
disrupted. Molecules that promote axon outgrowth (as for example
L1/NgCAM) still reach the axonal growth cone. This outgrowth is
slowed, though, compared to untreated neurons, because crucial
(currently unknown) axonal cargoes are slowed along the direct
axonal pathway. In addition, fusion of vesicles provides membrane
required for growth (Futerman and Banker, 1996). Therefore,
slowed transport along the direct pathway also leads to decreased
M.C. Chang et al. / Mol. Cell. Neurosci. 31 (2006) 525–538
534

Page 11

membrane supply for axon growth. The identification of endog-
enous axonal cargoes which use the direct pathway will, in the
future, illuminate the significance of this pathway.
Interestingly, disruption of lipid raft lipids has been shown to
delay surface transport in other cell types. Surface expression is
lowered by cholesterol depletion for GPI-linked proteins (Sheets et
al., 1997), for aminopeptidase in enterocytes (Hansen et al., 2000),
and for nicotinic acetylcholine receptor in CHO cells (Pediconi
et al., 2004). Changing cholesterol levels profoundly disturbs both
morphology of the Golgi (Hansen et al., 2000; Stuven et al., 2003;
Wang et al., 2000), as well as transport along multiple routes,
including clathrin-dependent endocytosis, caveolar endocytosis,
biogenesis of synaptic-like microvesicles, plasma membrane to
Golgi apparatus, TGN to apical surface, endosome to Golgi
transport, RE-mediated transport, and yeast vacuole fusion
(Chatterjee et al., 2001; Grimmer et al., 2000; Kato and Wickner,
2001; Thiele et al., 2000). ER to Golgi transport, on the other hand,
is not affected by changes in cholesterol content (Hansen et al.,
2000; Keller and Simons, 1998; Wang et al., 2000). Inhibiting
sphingolipid synthesis also affects transport. In the yeast S.
cerevisiae, inhibition of ceramide synthesis slowed the transport
of GPI-linked proteins, but not of soluble or transmembrane
proteins (Horvath et al., 1994). In CHO cells, the surface transport
of VSV-G (a basolaterally-expressed protein in epithelial cells) was
slowed 2–3-fold after treatment with PDMP, a sphingolipid
synthesis inhibitor (Rosenwald et al., 1992). This effect of PDMP
was not specific, but affected protein synthesis in general. In our
work, we observe no gross disturbance of all traffic or all
compartments. Rather, at the level of depletion achieved in our
cells, we could unmask the differential sensitivity of different kinds
of membrane cargoes to the lipid environment: only the transport
of NgCAM(CT3) is slowed.
L1 and several of its family members (neurofascin and
NrCAM) have been reported to associate with DRMs (Falk
et al., 2004; Ren and Bennett, 1998; Schafer et al., 2004).
Neurofascin is palmitoylated on a single cysteine residue within
the transmembrane domain and this palmitoylation might be
important for association with membrane lipids such as rafts (Ren
and Bennett, 1998). The DRM association of NgCAM is low in
our gradient system. Even though the degree of raft association
inside cells cannot be deduced from biochemical fractionation
behavior (Brown, 2001; London and Brown, 2000; Munro, 2003;
Shogomori and Brown, 2003), we can conclude that NgCAM
does appear to associate with DRMs less well than caveolin or
GM1 under the same conditions. NgCAM can therefore be
classified as a weakly DRM-associating protein by this criterion.
Could the DRM association of NgCAM be regulatable? Both
neurofascin and NrCAM are DRM-associated in a dynamic
fashion, rather than at all times. For instance, the association of
neurofascin155 with DRMs is developmentally regulated in glial
cells (Schafer et al., 2004). Similarly, the DRM association of
NrCAM is increased by cross-linking of its extracellular domain
(Falk et al., 2004). Work by others (Nakai and Kamiguchi, 2002)
has also demonstrated that L1 can be differentially associated
with DRMs at different stages of development: L1 association
with DRMs in cerebellum is only detectable between P8 and P28,
but not detectable at P56. The molecular basis for this regulated
DRM association is currently unknown. Our work additionally
shows that Triton-insolubility can be spatially regulated: A large
proportion of NgCAM(CT3) is Triton-insoluble in the Golgi
region, but not once it reaches the axonal plasma membrane. This
might be due to the subcellular distribution of presumptive
clustering molecules which recruit NgCAM(CT3) into rafts in the
Golgi, but are absent from the plasma membrane. The high
Triton-solubility of the axonal population also explains, at least in
part, why the overall DRM association on sucrose gradients is
relatively low. Since our work demonstrates that efficient
transport to the surface is sensitive to glycosphingolipid levels,
but fidelity of sorting is not, a proteinaceous machinery (which
might in itself be FB1-sensitive) is presumably responsible for
accurate axonal targeting on the direct pathway. GPI-linked
proteins, in contrast, appear to depend on GSLs for accurate
localization.
Axonal transport consists of multiple steps which are all
potentially subject to regulation (Winckler, 2004) and could be
differentially sensitive to FB1. The first step is efficient cargo
enrichment into a domain which will give rise to an axonal
carrier. Since NgCAM and NgCAM(CT3) have the same
extracellular and transmembrane domains, they do not differ in
their inherent capability to associate with DRMs (see Results) or
to bind a lumenal sorting component. They do differ in their
capacity to interact with cytoplasmic sorting components. We
showed previously that NgCAM contains crucial targeting signals
in its cytoplasmic tail (Wisco et al., 2003) which mediate
interaction with sorting adaptors to recruit vesicle coats (Anderson
et al., 2005). Without these cytoplasmic tail signals, surface
transport becomes less efficient (as seen for NgCAM(CT3)), but
occurs nevertheless with correct polarity. We propose that
receptors containing cytoplasmic sorting signals (like NgCAM
and LDLR) are sorted faithfully and efficiently to the axon (or
dendrites) even in the presence of FB1 due to interactions with a
cytoplasmic protein sorting machinery which mediates cargo
enrichment. In contrast, NgCAM lacking cytoplasmic sorting
signals is unable to directly bind to adaptors which promote
efficient inclusion into axonally-targeted carriers. Such tailless
molecules might be normally efficiently included into axonal
carriers because of lateral clustering. The presence of Triton-
insoluble NgCAM(CT3) in the Golgi region (see Results) is
consistent with this idea. Cargo enrichment could be enhanced by
cross-linking via lumenal raft-associated clustering agents. These
lumenal cross-linkers might recognize lumenal sorting determi-
nants (such as the domain identified in NgCAM by Sampo et al.,
2003). Interestingly, our findings differ from that of two axonal
proteins, Thy1 and PrPC, both of which are GPI-linked: the axonal
localization of both these proteins is disrupted by treatment with
FB1 (Galvan et al., 2005; Ledesma et al., 1998). Clearly, neurons
possess multiple pathways to the axon which are regulated
differently.
In summary, we report that the axonal targeting of NgCAM is
not sensitive to substantial depletion of ceramide-derived lipid
species. Similarly, NgCAM(CT3), a mutant NgCAM lacking the
cytoplasmic tail, is still axonally-targeted after FB1 treatment. We
therefore find no evidence that glycosphingolipids are necessary
for targeting of NgCAM. The effect of FB1, rather, is highly
specific: the kinetics of surface expression are slowed for the
truncated NgCAM(CT3). This is not due to a general slowing of all
Golgi-derived membrane traffic, since other molecules including
full-length NgCAM and LDLR are transported with normal
kinetics even in FB1-treated cells. Therefore, the effect of FB1
on the kinetics, but not the fidelity of a subset of axonal pathways,
can explain the observed inhibitory effect of FB1 on axonal
outgrowth.
M.C. Chang et al. / Mol. Cell. Neurosci. 31 (2006) 525–538
535

Page 12

Experimental methods
Reagents
The following antibodies were used: anti-L1 clone ASCS4 and anti-
NgCAM clone 8D9 from NIH hybridoma bank; anti-NgCAM antibody G4
from Peter Sonderegger (Universita ¨t Zu ¨rich) for Western blots; anti-LDLR
clone C7 from Ira Mellman (Yale University); anti-GRASP65 rabbit
antibody and anti-TGN38 rabbit antibody from Graham Warren (Yale
University); anti-transferrin receptor clone H68.4 from the ATCC; anti-
caveolin antibody from Santa Cruz Biochemicals; HRP-cholera toxin B and
Alexa488-cholera toxin B from Molecular Probes; Cy3-, Cy5-, and RRX-
conjugated and unconjugated Fab fragments and IgGs from Jackson
Immunologicals; Fumonisin B1 from Alexis Corporation.
Cell culture
Hippocampal cultures were prepared essentially as described (Winckler
et al., 1999). HEK293 cells were grown in DMEM with 10% FCS and 1%
glutamine for adenoviral production.
Adenoviral infection
Adenoviral infection and immunofluorescence were performed as
described (Wisco et al., 2003). Since we use the same antibody to detect
NgCAM and NgCAM(CT3), the range of staining intensities found is
comparable between different constructs. Images were captured with the
Orca cooled CCD camera (Hamamatsu) using Openlab software (Impro-
Vision) and processed identically in Adobe Photoshop. To visualize internal
and surface population of NgCAM separately, a protocol similar to
published procedures was used (Kamiguchi and Lemmon, 2000). This
protocol leads to separable staining of internal and external populations in
all cells (see Supplementary Fig. 2), except in very highly overexpressing
cells.
FB1 treatment
FB1 was dissolved in water at 2.5 mg/ml stock solution and stored
frozen. Cultures were treated with four equal doses of FB1 (final
concentration 25 AM) on alternate days starting on day 1 after plating, as
described in Ledesma et al., 1998. On day 8, cells were infected with
defective adenoviruses as described above and fixed at the indicated times
afterwards.
Preparation of detergent-resistant membranes (DRM)
Biochemical isolation of DRMs is a standard method for assessing the
potential of a protein to associate with lipid raft lipids. The assay for such a
preferential association has traditionally been to dissolve cells in cold
nonionic detergents and fractionate the lysate on sucrose gradients
(Chamberlain, 2004; Shogomori and Brown, 2003). Lipid raft lipids and
associated proteins tend to be insoluble in nonionic detergents under these
conditions and to float to a low density fraction in the gradient (Brown and
Rose, 1992). Caution needs to be taken in the interpretation of such
biochemical fractionation and controversy still exists as to the relation
between DRM association in vitro and lipid raft association in cells
(London and Brown, 2000). DRMs were prepared by a stringent procedure
essentially as previously described (Brown and Rose, 1992). Neurons were
grown for 10 days on 10 cm cell culture dishes, infected with adenovirus for
expression of the desired construct and harvested 48 h after infection. The
neurons were scraped off on ice in 1 ml TNEV (10 mM Tris–HCl, 150 mM
NaCl, 5 mM EDTA, pH 7.5) 1% (v/v) Triton X-100, Complete Protease
Inhibitor Cocktail (Roche), homogenized through repeated uptake in a 21
gauge needle, and incubated for 1 h at 4-C. The lysate was brought to
42.5%(w/v) sucrose using 85% sucrose in TNE with Triton X-100 and
loaded on the bottom of a linear gradient of 10 to 40% sucrose in TNE,
prepared using the Gradient Master (Biocomp). Gradients were centrifuged
for 19–22 h at 4-C at 39,000 rpm (200,000 ? g) in a Beckman SW41Ti
rotor. Fractions of 1 ml each were collected from the top. After TCA
precipitation, pellets were resuspended in Laemmli buffer and separated on
4–20% linear gradient SDS-PAGE gels (BioRad) and subsequently
transferred to nitrocellulose for Western analysis. The bottom two fractions
correspond to loaded cell lysates and include incompletely broken cells as
well as soluble material. No preclear-spin was performed and all material
was loaded onto the gradient. This allows for complete accounting, but
leads to a lower apparent DRM association due to the presence of unbroken
cells and aggregates at the bottom of the tube. The fractionation of
endogenously-expressed caveolin and cholera toxin-binding material (i.e.
GM1) were followed as positive controls to indicate where DRMs
fractionated under our conditions. Both caveolin and CT-B were found
both in the starting material at the bottom of the gradient as well as in lower
density fractions 5–8. Therefore, we consider fractions 5–8 to be DRM
fractions. In our gradients, about 50% of caveolin floated to DRM
fractions. Transferrin receptor is not associated with DRMs (Nakai and
Kamiguchi, 2002), and was used as a negative control.
Quantification of GM1
GM1 levels were quantified using either FITC-cholera toxin B or
HRP-cholera toxin B. Cultures were incubated after fixation with FITC-
cholera toxin B (1:6000 dilution). Images were taken with identical
camera settings and average pixel intensity along axons was quantified
using IPlab software along a one-pixel wide line.
Hippocampal neuronswere seeded onto two 3.5 cm cell culture dishes of
hippocampalneuronsandFumonisinB1was addedtooneofthe dishesfrom
the first day to a final concentration of 25 AM every second day as described
inLedesmaetal.(1998).Thecellswereharvestedonthe8thdayinculture(8
DIV) and harvested in TNEV buffer with 1% Triton X-100 and protease
inhibitorcocktail.Thelysateswerenormalizedtoequalproteincontentusing
a BCA protein assay (Pierce) and 2 Al of a dilution chain from 1:1 to 1:10?6
wasusedforgangliosidedetectionexperiments.Equalproteinconcentrations
were spotted as twofold dilutions onto nitrocellulose and incubated with
HRP-cholera toxin B for 1 h. After washing, the filter was developed in
Supersignal chemiluminescence reagent (Pierce) and exposed to X-ray film.
Western blots
The hippocampi of four E18 pups per plate were plated onto two 60 mm
tissue culture dishes coated with polylysine. FB1 treatment was done as
described above for one dish. On day 8, the cultures were infected with
AdNgCAM(CT3). After 34 h, cells were scraped into cold PBS with
Complete protease inhibitor cocktail (Boehringer). Protein concentration
was determined using the BCA Microkit (Pierce). Equal protein concen-
trations were loaded onto SDS-PAGE gels and transferred to nitrocellulose.
Blots were probed with anti-NgCAM antibody G4.
Triton X-100 insolubility assays for microscopy
Neurons were infected with AdNgCAM(CT3) for 28–30 h, and
extracted live in 0.5% Triton X-100 at either 4 or 37-C for 3 min. Cells
were subsequently fixed in 4% paraformaldehyde/3% sucrose/PBS and
stained with anti-NgCAM antibody 8D9. Non-extracted controls were
permeabilized after fixation. For quantification, pictures were takes at
identical settings and the average pixel intensity was determined in the
soma and on axons for 20–30 cells per experiment. Three independent
experiments were quantified. SEM is indicated in the bar graph.
Acknowledgments
We are grateful to Peter Sonderegger (University Zurich),
Graham Warren (Yale University), and Ira Mellman (Yale
M.C. Chang et al. / Mol. Cell. Neurosci. 31 (2006) 525–538
536

Page 13

University) for generously providing crucial reagents. We would
like to thank Drs. Heike Fo ¨lsch (Northwestern University), Serafin
Pinol-Roma (Sophie Davis School of Bio-Medical Education, City
College of New York), and Deanna Benson (Mount Sinai School
of Medicine MSSM) for critical comments on the manuscript and
Drs. Andreas Jenny (MSSM), Scott Henderson (Virginia Com-
monwealth University), Jeanne Hirsch (MSSM), and members of
the Winckler laboratory for helpful discussions. The MSSM
imaging facility is supported in part by grant NIH-NCI 1 R 24
CA095823-01. This work was supported by a Basil O’Connor
Scholarship (March of Dimes Foundation), and NINDS NIH
1RO1NS045969 (both to BW).
Appendix A. Supplementary data
Supplementary data associated with this article can be found in
the online version at doi:10.1016/j.mcn.2005.11.006.
References
Anderson, E., Maday, S., Sfakianos, J., Hull, M., Winckler, B., Sheff, D.,
Fo ¨lsch, H., Mellman, I., 2005. Transcytosis of NgCAM in epithelial
cells reflects differential signal recognition on the endocytic and
secretory pathways. J. Cell Biol. 170, 595–605.
Brown, D.A., 2001. Seeing is believing: visualization of rafts in model
membranes. Proc. Natl. Acad. Sci. U. S. A. 98, 10517–10518.
Brown, D.A., Rose, J.K., 1992. Sorting of GPI-anchored proteins to
glycolipid-enriched membrane subdomains during transport to the
apical cell surface. Cell 68, 533–544.
Chamberlain, L.H., 2004. Detergents as tools for the purification and
classification of lipid rafts. FEBS Lett. 559, 1–5.
Chatterjee, S., Smith, E.R., Hanada, K., Stevens, V.L., Mayor, S., 2001.
GPI anchoring leads to sphingolipid-dependent retention of endocy-
tosed proteins in the recycling endosomal compartment. EMBO J. 20,
1583–1592.
Craig, A.M., Banker, G., 1994. Neuronal polarity. Annu. Rev. Neurosci. 17,
267–310.
El-Husseini, A.-D., Craven, S.E., Brock, S.C., Bredt, D.S., 2001. Polarized
targeting of peripheral membrane proteins in neurons. J. Biol. Chem.
276, 44984–44992.
Falk, J., Thoumine, O., Dequidt, C., Choquet, D., Faivre-Sarrailh, C.,
2004. NrCAM coupling to the cytoskeleton depends on multiple
protein domains and partitioning into lipid rafts. Mol. Biol. Cell 15,
4695–4709.
Fullekrug, J., Simons, K., 2004. Lipid rafts and apical membrane traffic.
Ann. N. Y. Acad. Sci. 1014, 164–169.
Futerman, A.H., Banker, G.A., 1996. The economics of neurite out-
growth—The addition of new membrane to growing axons. Trends
Neurosci. 19, 144–149.
Futerman, A.H., Riezman, H., 2005. The ins and outs of sphingolipid
synthesis. Trends Cell Biol. 15, 312–318.
Galvan, C., Camoletto, P.G., Dotti, C.G., Aguzzi, A., Ledesma, M.D., 2005.
Proper axonal distribution of PrPC depends on cholesterolsphingomye-
lin-enriched membrane domains and is developmentally regulated in
hippocampal neurons. Mol. Cell Neurosci. 30, 304.
Ginzburg, L., Kacher, Y., Futerman, A.H., 2004. The pathogenesis of gly-
cosphingolipid storage disorders. Semin. Cell Dev. Biol. 15, 417–431.
Grimmer, S., Iversen, T.G., van Deurs, B., Sandvig, K., 2000. Endosome to
Golgi transport of ricin is regulated by cholesterol. Mol. Biol. Cell 11,
4205–4216.
Guirland, C., Suzuki, S., Kojima, M., Lu, B., Zheng, J.Q., 2004. Lipid
rafts mediate chemotropic guidance of nerve growth cones. Neuron
42, 51–62.
Hansen, G.H., Niels-Christiansen, L.L., Thorsen, E., Immerdal, L.,
Danielsen, E.M., 2000. Cholesterol depletion of enterocytes. Effect on
the Golgi complex and apical membrane trafficking. J. Biol. Chem. 275,
5136–5142.
Harder, T., 2004. Lipid raft domains and protein networks in T-cell receptor
signal transduction. Curr. Opin. Immunol. 16, 353–359.
Harel, R., Futerman, A.H., 1993. Inhibition of sphingolipid synthesis
affects axonal outgrowth in cultured hippocampal neurons. J. Biol.
Chem. 268, 14476–14481.
Helms, J.B., Zurzolo, C., 2004. Lipids as targeting signals: lipid rafts and
intracellular trafficking. Traffic 5, 247–254.
Holthuis, J.C., van Meer, G., Huitema, K., 2003. Lipid microdomains, lipid
translocation and the organization of intracellular membrane transport.
Mol. Membr. Biol. 20, 231–241.
Horton, A.C., Ehlers, M.D., 2003. Neuronal polarity and trafficking.
Neuron 40, 277–295.
Horvath, A., Sutterlin, C., Manning-Krieg, U., Movva, N.R., Riezman, H.,
1994. Ceramide synthesis enhances transport of GPI-anchored proteins
to the Golgi apparatus in yeast. EMBO J. 13, 3687–3695.
Ikonen, E., 2001. Roles of lipid rafts in membrane transport. Curr. Opin.
Cell Biol. 13, 470–477.
Jacob, R., Naim, H.Y., 2001. Apical membrane proteins are transported in
distinct vesicular carriers. Curr. Biol. 11, 1444–1450.
Jareb, M., Banker, G., 1998. The polarized sorting of membrane proteins
expressed in cultured hippocampal neurons using viral vectors. Neuron
20, 855–867.
Kamiguchi, H., Lemmon, V., 2000. Recycling of the cell adhesion molecule
L1 in axonal growth cones. J. Neurosci. 20, 3676–3686.
Kato, M., Wickner, W., 2001. Ergosterol is required for the Sec18/ATP-
dependent priming step of homotypic vacuole fusion. EMBO J. 20,
4035–4040.
Keller, P., Simons, K., 1998. Cholesterol is required for surface transport of
influenza virus hemagglutinin. J. Cell Biol. 23, 1357–1367.
Lamaze, C., Dujeancourt, A., Baba, T., Lo, C.G., Benmerah, A., Dautry-
Varsat, A., 2001. Interleukin 2 receptors and detergent-resistant
membrane domains define a clathrin-independent endocytic pathway.
Mol. Cell 7, 661–671.
Ledesma, M.D., Simons, K., Dotti, C.G., 1998. Neuronal polarity: essential
role of protein–lipid complexes in axonal sorting. Proc. Natl. Acad. Sci.
U. S. A. 95, 3966–3971.
Ledesma, M.D., Brugger, B., Bunning, C., Wieland, F.T., Dotti, C.G., 1999.
Maturation of the axonal plasma membrane requires upregulation of
sphingomyelin synthesis and formation of protein–lipid complexes.
EMBO J. 18, 1761–1771.
Lipardi, C., Nitsch, L., Zurzolo, C., 2000. Detergent-insoluble GPI-
anchored proteins are apically sorted in fischer rat thyroid cells, but
interference with cholesterol or sphingolipids differentially affects
detergent insolubility and apical sorting. Mol. Biol. Cell 11, 531–542.
Lisanti, M.P., Rodriguez-Boulan, E., 1990. Glycophospholipid membrane
anchoring provides clues to the mechanism of protein sorting in
polarized epithelial cells. Trends Biochem. 15, 113–116.
London, E., Brown, D.A., 2000. Insolubility of lipids in triton X-100:
physical origin and relationship to sphingolipid/cholesterol membrane
domains (rafts). Biochim. Biophys. Acta 1508, 182–195.
Mays, R.W., Siemers, K.A., Fritz, B.A., Lowe, A.W., van Meer, G., Nelson,
W.J., 1995. Hierarchy of mechanisms involved in generating
Na+K+ATPase polarity in MDCK epithelial cells. J. Cell Biol. 130,
1105–1115.
Mellman, I., 1995. Molecular sorting of membrane proteins in polarized
and nonpolarized cells. Cold Spring Harbor Symp. Quant. Biol. LX,
745–753.
Munro, S., 2003. Lipid rafts: elusive or illusive? Cell 115, 377–388.
Nakai, Y., Kamiguchi, H., 2002. Migration of nerve growth cones requires
detergent-resistant membranes in a spatially defined and substrate-
dependent manner. J. Cell Biol. 159, 1097–1108.
Nelson, W.J., Yeaman, C., 2001. Protein trafficking in the exocytic pathway
of polarized epithelial cells. Trends Cell Biol. 11, 451–483.
M.C. Chang et al. / Mol. Cell. Neurosci. 31 (2006) 525–538
537

Page 14

Paladino, S., Sarnataro, D., Pillich, R., Tivodar, S., Nitsch, L., Zurzolo, C.,
2004. Protein oligomerization modulates raft partitioning and apical
sorting of GPI-anchored proteins. J. Cell Biol. 167, 699–704.
Pediconi, M.F., Gallegos, C.E., De Los Santos, E.B., Barrantes, F.J., 2004.
Metabolic cholesterol depletion hinders cell-surface trafficking of the
nicotinic acetylcholine receptor. Neuroscience 128, 239–249.
Peretti, D., Peris, L., Rosso, S., Quiroga, S., Caceres, A., 2000. Evidence
for the involvement of KIF4 in the anterograde transport of L1-
containing vesicles. J. Cell Biol. 149, 141–152.
Polishchuk, R., Di Pentima, A., Lippincott-Schwartz, J., 2004. Delivery of
raft-associated, GPI-anchored proteins to the apical surface of polarized
MDCK cells by a transcytotic pathway. Nat. Cell Biol. 6, 297–307.
Rajendran, L., Simons, K., 2005. Lipid rafts and membrane dynamics.
J. Cell Sci. 118, 1099–1102.
Ren, Q., Bennett, V., 1998. Palmitoylation of neurofascin at a site in the
membrane-spanning domain highly conserved among the L1 family of
cell adhesion molecules. J. Neurochem. 70, 1839–1849.
Rodriguez-Boulan, E., Kreitzer, G., Musch, A., 2005. Organization of
vesicular trafficking in epithelia. Nat. Rev., Mol. Cell Biol. 6, 233–247.
Rosenwald, A.G., Machamer, C.E., Pagano, R.E., 1992. Effects of a
sphingolipid synthesis inhibitor on membrane transport through the
secretory pathway. Biochemistry 31, 3581.
Sampo, B., Kaech, S., Kunz, S., Banker, G., 2003. Two distinct
mechanisms target membrane proteins to the axonal surface. Neuron
37, 611–624.
Sarnataro, D., Paladino, S., Campana, V., Grassi, J., Nitsch, L., Zurzolo, C.,
2002. PrPC is sorted to the basolateral membrane of epithelial cells
independently of its association with rafts. Traffic 3, 810–821.
Schafer, D.P., Bansal, R., Hedstrom, K.L., Pfeiffer, S.E., Rasband, M.N.,
2004. Does paranode formation and maintenance require partitioning of
neurofascin 155 into lipid rafts? J. Neurosci. 24, 3176–3185.
Schuck, S., Simons, K., 2004. Polarized sorting in epithelial cells: raft
clustering and the biogenesis of the apical membrane. J. Cell Sci. 117,
5955–5964.
Schwarz, A., Futerman, A.H., 1997. Distinct roles for ceramide and
glucosylceramide at different stages of neuronal growth. J. Neurosci.
17, 2929.
Schwarz, A., Rapaport, E., Hirschberg, K., Futerman, A.H., 1995. A
regulatory role for sphingolipids in neuronal growth. Inhibition of
sphingolipid synthesis and degradation have opposite effects on axonal
branching. J. Biol. Chem. 270, 10990–10998.
Sheets, E.D., Lee, G.M., Simson, R., Jacobson, K., 1997. Transient
confinement of a glycosylphosphatidylinositol-anchored protein in the
plasma membrane. Biochemistry 36, 12449–12458.
Shogomori, H., Brown, D.A., 2003. Use of detergents to study membrane
rafts: the good, the bad, and the ugly. Biol. Chem. 384, 1259–1263.
Shogomori, H., Futerman, A.H., 2001. Cholera toxin is found in detergent-
insoluble rafts/domains at the cell surface of hippocampal neurons but is
internalized via a raft-independent mechanism. J. Biol. Chem. 276,
9182–9188.
Shogomori, H., Burack, M.A., Banker, G., Futerman, A.H., 1999. Ganglio-
sides GM1 and GD1b are not polarized in mature hippocampal neurons.
FEBS Lett. 458, 107–111.
Simons, K., Ikonen, E., 1997. Functional rafts in cell membranes. Nature
387, 569–572.
Stuven, E., Porat, A., Shimron, F., Fass, E., Kaloyanova, D., Brugger, B.,
Wieland, F.T., Elazar, Z., Helms, J.B., 2003. Intra-Golgi protein
transport depends on a cholesterol balance in the lipid membrane.
J. Biol. Chem. 278, 53112–53122.
Thiele, C., Hannah, M.J., Fahrenholz, F., Huttner, W.B., 2000. Cholesterol
binds to synaptophysin and is required for biogenesis of synaptic
vesicles. Nat. Cell Biol. 2, 42–49.
Tuma, P.L., Hubbard, A.L., 2003. Transcytosis: crossing cellular barriers.
Physiol. Rev. 83, 871.
Vogt, L., Giger, R.J., Ziegler, U., Kunz, B., Buchstaller, A., Hermens,
W.T.J.M.C., Kaplitt, M.G., Rosenfeld, M.R., Pfaff, D.W., Verhaagen, J.,
Sonderegger, P., 1996. Continuous renewal of the axonal pathway
sensor apparatus by insertion of new sensor molecules into the growth
cone membrane. Curr. Biol. 6, 1153–1158.
Wang, Y., Thiele, C., Huttner, W.B., 2000. Cholesterol is required for the
formation of regulated and constitutive secretory vesicles from the
trans-Golgi network. Traffic 1, 952–962.
Winckler, B., 2004. Pathways for axonal targeting of membrane proteins.
Biol. Cell 96, 669–674.
Winckler, B., Mellman, I., 1999. Neuronal polarity: controlling the sorting
and diffusion of membrane components. Neuron 23, 637–640.
Winckler, B., Forscher, P., Mellman, I., 1999. A diffusion barrier maintains
distribution of membrane proteins in polarized neurons. Nature 397,
698–701.
Wisco, D., Anderson, E.D., Chang, M.C., Norden, C., Boiko, T., Folsch, H.,
Winckler, B., 2003. Uncovering multiple axonal targeting pathways in
hippocampal neurons. J. Cell Biol. 162, 1317–1328.
M.C. Chang et al. / Mol. Cell. Neurosci. 31 (2006) 525–538
538