Dissociation of ERK signalling inhibition from the anti-amyloidogenic action of synthetic ceramide analogues

Amyloid plaques are a key pathological feature of AD (Alzheimer's disease) [1], a common neurodegenerative disorder affecting our aging population. Amyloid plaques are enriched with Aβ (amyloid β-peptide) peptides that are derived from the proteolytic processing of the APP (amyloid precursor protein) [2]. APP can be proteolytically processed by the α-secretase pathway that does not generate Aβ as the metalloproteases [such as ADAM-10 (a disintegrin and metalloproteinase 10)] responsible cleave in the middle of the Aβ sequence. In an alternative pathway, sequential cleavage of APP by β-secretase [BACE-1 (β-site APP-cleaving enzyme)] and γ-secretase [a complex containing PS-1 (presenilin 1) or PS-2 as the catalytic subunit] generates neurotoxic Aβ peptides predominantly of 40 or 42 amino acids [3–6]. In addition to factors that regulate the rate of Aβ production, the ratio of Aβ40 [Aβ-(1–40)] to Aβ42 [Aβ-(1–42)] species generated, their propensity to form macromolecular complexes and their catabolism and clearance from the central nervous system are all relevant therapeutic targets for AD [3,7]. Despite the recognized role for Aβ in AD neurodegeneration, the factors that modify Aβ production and deposition are not completely understood.

Previous studies have suggested that specific GSLs (glycosphingolipids) may promote Aβ production and/or its assembly into neurotoxic complexes [8–13]. GSLs co-localize with Aβ in amyloid plaques and it has been proposed that ganglioside G_M1 interacts with Aβ to form a seed for amyloid plaque formation [14]. Other studies have shown that G_M1 can act on mature Aβ fibrils in amyloid plaque to regenerate neurotoxic Aβ protofibrils [15]. In addition, when GD3 synthase-knockout mice (phenotype characterized by reductions in the levels of several brain gangliosides) were crossed with APPswe+PSEN1DE9 amyloidogenic mice (a mouse model of AD displaying both amyloid plaque pathology and cognitive impairments), both soluble Aβ and plaque load were reduced (85–95%) and this was associated with improved cognitive performance [16]. These data imply that pharmacological inhibition of GSL synthesis may be beneficial in the AD/amyloidogenesis context.

A previous study has shown that the GSL synthesis inhibitor PDMP (D-1-phenyl-2-decanoylamino-3-morpholino-1-propanol) reduced Aβ secretion from SH-SY5Y neuroblastoma cells by ~55% when used at a concentration of 25 μM for a 48 h period [10]. In recent studies, we have shown that two more potent GSL synthesis inhibitors, PPMP (D-threo-1-phenyl2-hexadecanoylamino-3-morpholino-1-propanol) and EtDO-P4 (D-threo-ethylenedioxy-1-phenyl-2-palmitoylamino-3-pyrrolidino-1-propanol), inhibit Aβ secretion both from CHO cells (Chinese-hamster ovary cells) stably expressing human APP695 (CHO-APP) and from primary human neurons with IC₅₀ values in the low (1–5) micromolar range [17]. Although the D-threo-enantiomers of these ceramide analogues have been shown to have a high degree of specificity for GCS (glucosylceramide synthase) [18–23], downstream effects on membrane lipid microdomain structure, cellular ceramide metabolism and signalling pathways may also be involved. For example, modulation of membrane GSL levels by PDMP and related compounds has an impact on membrane cholesterol transport and EGFR (epidermal growth factor receptor) signalling [24,25]. PDMP has also been shown to inhibit ERK (extracellular-signal-regulated kinase) phosphorylation in neural precursor cells [26] and this is potentially important in the current context as there is evidence that the ERK pathway is dysregulated in AD [27–29] and that ERK activation may regulate APP proteolysis and/or Aβ production in vitro [30–32].

In the present study, we have investigated the enantiomer-specific actions of PPMP and EtDO-P4 on both cellular Aβ production and ERK signalling. We also directly assess the impact that ERK inhibitors and siRNA (small interfering RNA)-mediated silencing of GCS has on cellular Aβ production.

Synthetic ceramide analogues PDMP, D-PPMP and L-PPMP were purchased from Matreya. D/L-EtDO-P4 was synthesized as described previously [20]. The D- and L-EtDO-P4 enantiomers were purified by preparative normal phase HPLC using a Lux 5 μm Cellulose-2 AXIA packed chiral 250 mm×21.2 mm column (Phenomenexa). An Agilent 1100 HPLC system was used with a mobile phase hexane/propan-2-ol/diethylamine (85:15:0.1, by vol.), a flow rate of 10 ml/min and UV detection at 220 nm. D-EtDO-P4 eluted at 25 min and was collected from the column and dried under vacuum before use in experiments. The γ-secretase inhibitor DAPT (N-[N-(3,5-difluorophenacetyl)-L-alanyl]-(S)-phenylglycine tbutyl ester) and the MAPK (mitogen-activated protein kinase) pathway inhibitors PD98059 (2-(2-amino-3-methoxyphenyl)-4H-1-benzopyran-4-one) and U0126 (1,4-diamino-2,3-dicyano-1,4-bis[2-aminophenylthio] butadiene) were purchased from Sigma. Purified leech (Macrobdella decora) ceramide glycanase (EC 3.2.1.123) was from V-Labs. Cell culture medium and additives were obtained from Invitrogen unless stated otherwise. All organic solvents were of analytical grade and were purchased from Ajax Finechem. All other reagents were purchased through standard commercial suppliers.

The CHO cell line stably expressing the human 695-amino-acid isoform of APP (CHO-APP) was generated and maintained as described previously [33,34]. CHO-APP cells were cultured in RPMI 1640 medium containing 10% (v/v) FBS (fetal bovine serum), 2 mM glutamine, 100 units/ml penicillin and 100 μg/ml streptomycin. The cells were cultured in the presence of puromycin (7.5 μg/ml) to maintain selection for the expression plasmid. The SH-SY5Y-APP cell line was generated by transfecting human SH-SY5Y neuroblastoma cells with cDNA containing wild-type human APP695 cloned into the pIRESpuro2 vector [35]. Stably transfected cells were selected using 2 μg/ml puromycin and cultured in RPMI 1640 medium containing 20% (v/v) FBS, 2 mM glutamine, 100 units/ml penicillin, 100 μg/ml streptomycin and 2 μg/ml puromycin to maintain selection for the expression plasmid. SH-SY5Y-APP cells were differentiated to generate a neuronal phenotype using 10 μM all-trans-retinoic acid (R2625; Sigma) for 7 days as previously described [36]. Cells were treated with GSL inhibitors, MAPK inhibitors or DAPT as indicated for up to 48 h. These compounds were added to cells in complete growth medium containing 10% (v/v) FBS.

Cells were grown in 96-well plates (5000 cells/well) and maintained as described above. Cytotoxicity was assessed by incubating cells in 100 μl of medium containing various concentrations of GSL inhibitors as indicated along with vehicle controls (six wells per concentration). After 48 h the culture medium was removed, and 100 μl of medium [DMEM (Dulbecco's modified Eagle's medium) and 10% (v/v) FBS] containing 0.5 mg/ml MTT (M2128; Sigma) was added and the cells were incubated for 1 h at 37°C. The medium was discarded, and the cells were dissolved in DMSO (100 μl per well) and the absorbance of the cell lysates was measured at 550 nm. Higher absorbace values indicate increased cell viability.

Cell culture medium was collected from CHO-APP and SH-SY5Y-APP cells and pre-diluted into PBS at 1:100 for Aβ40 and at 1:10 for Aβ42 in preparation for Aβ measurement by ELISA. βMark X-40 and X-42 ELISA kits (SIG-38956-Kit and SIG-38954-Kit; Covance) were used to quantify Aβ concentration, according to the manufacturer's instructions as described previously [37].

Cells were routinely cultured in 12-well plates, rinsed with ice-cold PBS and lysed in lysis buffer (20 mM Tris/HCl, pH 7.5, 150 mM NaCl, 1 mM EDTA, 1% Nonidet P40, 0.5% deoxycholate, 0.1% SDS and protease and phosphatase inhibitors). Bicinchoninic acid protein assays were performed on lysates and equal amounts of protein were separated by SDS/PAGE (12% gels) and transferred on to 0.2 μm nitrocellulose membranes at 70 V for 20 min. Membranes were blocked at 22°C for 2 h in PBS containing non-fat dried skimmed milk powder and probed with the relevant antibodies at 4°C for 16 h to analyse APP using W0-2 monoclonal antibody [34]. The W0-2 antibody recognizes an epitope between residues 2 and 8 at the N-terminal of the Aβ peptide sequence [38,39]. ERK signalling was assessed by Western blotting using phospho-p44/42 MAPK [pERK1/2 (phosphorylated ERK 1/2)] and p44/42 MAPK (ERK1/2) rabbit monoclonal antibodies at 1:5000 dilution (#4377 and #4695 respectively; Cell Signaling Technology). Where indicated, the membranes were stripped and re-probed for β-actin (rabbit polyclonal 1:5000; A5060; Sigma), which was used as a loading control. Western blotting of secreted sAPPα (soluble APPα fragment) and secreted Aβ peptides was carried out as described previously using W0-2 monoclonal antibody [34]. Both sAPPα and Aβ were routinely analysed in growth medium containing 10% (v/v) FBS after 48 h incubation either with or without the treatments indicated. Signals were detected using ECL® (enhanced chemiluminescence; GE Healthcare) and X-ray films. The signal intensity was quantified using NIH ImageJ software and the data presented as a ratio relative to non-treated cells as described previously [34].

RNA was isolated from cells using TRI Reagent (T9424; Sigma) following the manufacturer's protocol. All procedures were carried out using RNase-free reagents and quantification of total RNA was done by using a NanoDrop ND-1000 spectrophotometer. RNA (5 μg) was reverse transcribed into cDNA as described previously [34,40]. Conventional PCR amplification was carried out with 30 cycles of denaturation (95°C, 30 s), annealing (58°C, 30 s) and extension (72°C, 30 s), and the PCR products were visualized after electrophoresis in 1% agarose gels. The primers used are the following: GCS (GenBank® accession number AF351131.1; cDNA sequence, 141–320, 180 bp), forward, 5′-TAGCAAGCTCCCTGGTGTCT-3′; reverse, 5′-TTCGGGTATTTTCCAAGCAG-3′; and GAPDH (glyceraldehyde-3-phosphate dehydrogenase; GenBank® accession number X52123.1; cDNA sequence, 309–456, 148 bp), forward, 5′-GGAGAAGGCTGGGGCCCACT-3′; reverse, 5′-GGTGGTGCAGGACGCATTGCT-3′.

Transfection of siRNA into CHO-APP cells was performed using commercially prepared siRNA duplexes (Invitrogen Stealth™). The sequence targeted by the siRNA was selected using the BLOCK-iT RNAi designer program (www.invitrogen.com/rnai), which was found to be in close agreement with available published sequences [41–43]. RNA sequences were 5′-UUAAACUGGCAACAAAGCAUUCUGA-3′, and 5′-UCAGAAUGCUUUGUUGCCAGUUUAA-3′ (Invitrogen synthesis number AF351131_stealth_880). The siRNA was targeted to GCS (UDP-glucose ceramide glucosyltransferase, UGCG; GenBank® accession number AF351131.1; cDNA sequence, 880–824)-specific sequence 5′-TCACGTAGTTTCGTTTGACTAGTAA-3′ and a random reorganization of this sequence was used as the control siRNA 5′-GAATATGCGTTTATACCGTGTTACT-3′. siRNA duplexes were transfected by using Lipofectamine™ RNAiMAX (13778-075; Invitrogen) by a forward transfection procedure in six-well culture plates following the manufacturer's protocol. Transfection efficiency was confirmed by substituting the siRNA with fluorescently labelled BLOCK-iT reagent, supplied with the kit, using the same transfection procedure and this confirmed >95% transfection efficiency was achieved as assessed by fluorescence microscopy (results not shown).

Cell supernatants and lysates were collected and processed for Western blot analysis as above. RNA extraction, cDNA synthesis and quantitative real-time PCR was performed as described previously [34,40]. Briefly, real-time PCR was carried out using an Eppendorf Mastercycler® ep realplex Thermal Cycler using the fluorescent dye SYBR Green (Eppendorf). Amplification was carried out with 40 cycles of 94°C for 15 s and 60°C for 1 min. All gene expression was normalized to GAPDH, which served as an internal control for the quality of RNA isolated from each cell sample. Experiments were performed in triplicate and at least three samples were analysed for each treatment. The level of expression for each gene was calculated using the comparative threshold cycle value (C_t) value method using the formula 2^−ΔΔCt (where ΔΔC_t=ΔC_t sample−ΔC_t reference) as described previously [44,45].

GSL profiles of CHO-APP cells were analysed as described previously [24]. Briefly, cells grown to confluence in six-well plates were treated with synthetic ceramide analogues as described above, rinsed three times with PBS and lipids extracted in 2:1 (v/v) chloroform/methanol. The GSL fractions were isolated by silicic acid chromatography, and the glycan moiety was released by ceramide glycanase addition [46]. The GSL glycans were then fluorescently labelled with 2AA (2-anthranilic acid) and analysed by normal phase HPLC as described previously [24]. Total peak area units for the glycan peaks were pooled to calculate the reduction of cellular GSL levels after treatment with ceramide analogues or GCS siRNA. We noted that interfering peaks eluting in the positions of glucose and lactose were detected in the commercial ceramide glycanase enzyme blank conditions, and levels of GlcCer (glucosylceramide) and LacCer (lactosylceramide) were therefore excluded from the analyses of CHO-APP GSL levels. The principal GSLs analysed (asialo ganglioside G_M2, globotriaosylceramide, G_M3 and globotetraosylceramide) were identified based on their glucose unit values by comparison with a 2-AA-labelled dextran standard as previously described in detail [46,47]. Values were expressed as a percentage of total GSL levels present in mock-treated CHO-APP cells.

Unless stated otherwise, experiments were performed in triplicate and with three experimental replicates. Results are presented as means with S.E.M. shown by error bars. Differences were considered significant where P<0.05, as determined using a two-tailed Student's t test for unpaired data.

Previous studies have shown that PDMP inhibits Aβ production from the human SH-SY5Y neuronal cell line by ~55% when used at a concentration of 25 μM for 48 h [10]. We recently confirmed this finding and showed that D-PDMP also inhibits Aβ production in the well-characterized CHO-APP cell line [17]. In the latter experiments, 48 h treatment of CHO-APP cells with 15 μM D-PDMP reduced Aβ levels by ~50% and an IC₅₀ value of 15.8 μM was calculated for PDMP-mediated inhibition of Aβ production [17]. Using identical experimental conditions, we assessed whether ERK phosphorylation may be modulated by D-PDMP. At a concentration of 15 μM, D-PDMP inhibited Aβ production by 44% as predicted (Figure 1). Under these conditions, pERK levels were also reduced by 27% with no change in total ERK (Figure 1). Since our previous studies indicated that the structurally related PDMP analogues, PPMP and EtDO-P4, were more potent anti-amyloidogenic compounds that were more likely to be useful as prototype therapeutic agents in vivo [17], we next assessed the impact of these compounds on cellular Aβ production and ERK activation.

Earlier studies indicated that the D-threo-enantiomers of PDMP and structurally related compounds represent the active forms of these GSL synthesis inhibitors [18–23]. We therefore assessed the enantiomer-specific effects of PPMP and EtDO-P4 on both Aβ production and ERK activation. Consistent with previous data [17], 5 μM D-PPMP inhibited Aβ production by 49%, whereas L-PPMP had no effect at the same concentration (Figure 2). Cellular APP and secreted sAPPα were unaffected by these compounds (Figure 2). In contrast, both D- and L-PPMP significantly inhibited ERK phosphorylation (Figure 2). Similarly, when we assessed the D- and L-threo enantiomers of EtDO-P4 at a concentration of 1 μM, only the D-EtDO-P4 (which was the more potent GSL synthesis inhibitor) significantly inhibited Aβ production, whereas both D- and L-EtDO-P4 inhibited ERK phosphorylation (Figure 3). Unexpectedly, the L-EtDO-P4 that we separated from the D/L-EtDO-P4 mixture by preparative chiral HPLC (results not shown) also reduced cellular GSL levels by 43% (Figure 3A). Even though the inhibition of Aβ production by the D-enantiomers of all three ceramide analogues (PDMP, PPMP and EtDO-P4) occurred concomitantly with inhibition of ERK phosphorylation, this is unlikely to represent their primary mode of anti-amyloidogenic action since the L-enantiomers also inhibited ERK activation without significantly reducing Aβ production. It was also intriguing that the L-EtDO-P4-mediated reductions in cellular GSL levels (which were less pronounced than the GSL reduction resulting from D-EtDO-P4) did not significantly reduce Aβ production.

To rule out the possibility that the enantiomer-specific effect of PPMP on cellular Aβ production was simply due to the concentration used, various concentrations of both D-PPMP and L-PPMP were tested in parallel. Since these compounds can become cytotoxic at high concentrations, we first determined cell viability at various inhibitor doses using the MTT assay. Consistent with our previous data [17], cells remained ~85% viable after 48 h treatment with 5 μM D-PPMP (Figure 4A). Increased toxicity was detected when the concentration of the inhibitor was raised to 10 μM or above (Figure 4A). At PPMP concentrations above 10 μM, cells also began to detach from the plates and therefore data pertaining to MTT reduction and Aβ levels at 15 μM are not quantitative and are therefore not presented herein. The cytotoxicity profiles of both the D- and L-isomers in the 1–10 μM range were very similar (Figure 4A). Importantly, D-PPMP dose-dependently inhibited Aβ generation, whereas L-PPMP did not (Figure 4). The levels of another W0-2-positive protein of unknown identity at a molecular mass of ~15 kDa did not change with inhibitor treatment in the 1–10 μM range, suggesting that the impact of D-PPMP is not a generalized action on cellular protein production.

We next investigated the enantiomer-specific inhibition of cellular Aβ production in the SH-SY5Y-APP neuronal cell line. Both D-PPMP and L-PPMP displayed similar cytotoxicity profiles in the neuronal cells as compared with the CHO-APP cells (Figure 5A; cf Figure 4A). In the neuronal cell line viability remained at ~90% at inhibitor concentrations of 5 μM (Figure 5A). Furthermore, only the D-enantiomer inhibited Aβ production, whereas both the D- and L-enantiomers inhibited ERK phosphorylation (Figure 5). In order to determine if D-PPMP selectively inhibits either Aβ40 or Aβ42 production, we used an ELISA to quantify these species. This indicated that the CHO-APP cell line generates 7.5-fold more Aβ40 and 8.1-fold more Aβ42 than the SH-SY5Y cell lines and that in both cases the amount of Aβ40 produced was ~15-fold higher than Aβ42 (Figure 6). Subsequent to treatment with 5 μM D-PPMP, production of Aβ40 was reduced by 74 and 81% in the CHO-APP and SH-SY5Y-APP cell lines respectively, whereas Aβ42 was reduced by 38% but only in CHO-APP cells (Figure 6). This indicates that the reduction in Aβ is predominantly due to reduction in Aβ40. There were no significant reductions in Aβ when cells were treated with L-PPMP and the ELISA data are therefore in good agreement with the semi-quantitative Western blot analysis.

Taken together, these experiments raise two important issues regarding: (i) the direct role that ERK activation may play in amyloidogenic APP processing and (ii) the specificity of reduced cellular GSL synthesis as a modulator of Aβ production. These issues were addressed using alternative approaches to inhibit ERK phosphorylation or GSL synthesis. To probe for a direct role that ERK signalling may play in Aβ production, CHO-APP cells were treated with two different well-characterized MAPK pathway inhibitors, PD98059 or U0126. Both PD98059 and U0126 strongly inhibited pERK formation; however, Aβ production was not suppressed (Figure 7). In fact, for reasons that remain unknown, PD98059 stimulated Aβ production by 20% (Figure 7). The results from these experiments further suggest that the inhibition of ERK activation by PDMP, PPMP and EtDO-P4 is not likely to underlie their actions as inhibitors of cellular Aβ production.

In a final set of experiments we assessed the impact that inhibition of GSL synthesis using siRNA has on both cellular Aβ production and ERK phosphorylation. Targeting of GCS using a 40 nM concentration of siRNA duplex for 48 h reduced GCS mRNA levels by 82% (Figure 8). This resulted in a 77% decrease in CHO-APP cellular GSL levels (Figure 8). The relative reductions in the levels of the different GSL species were similar when either ceramide analogues (see Supplementary Figure S1 at http://www.clinsci.org/cs/122/cs1220409add.htm) or siRNA (Figure 8) was used to inhibit GCS. In parallel experiments, GCS siRNA treatment of CHO-APP cells did not result in any detectable changes in either Aβ production or ERK phosphorylation (Figure 9). These data indicate that reductions in cellular GSL levels alone are not sufficient to reduce cellular Aβ production.

There is accumulating data to suggest that alterations in cerebral lipid homoeostasis may be directly associated with AD (see [48] for a review). Based on the fact that APP is a type I membrane protein that is proteolytically processed to form Aβ in sphingolipid-enriched membrane microdomains, the possibility that modulation of membrane lipid composition could represent a novel means to regulate Aβ production presents a plausible basis for AD therapeutic intervention [9,11,49–52].

Studies from different laboratories have shown that the ceramide analogue inhibitors of GCS that are based on the PDMP structure inhibit Aβ production in vitro [10,17]; however, the mechanism(s) of their anti-amyloidogenic action remain to be fully elucidated. In the present studies, we have shown that the inhibition of ERK phosphorylation mediated by these GSL inhibitors is not likely to contribute to their anti-amyloidogenic actions since both D- and L-enantiomers inhibit ERK activation, whereas only the D-enantiomers suppress cellular Aβ production. It remains possible that the efficacy of additional anti-amyloidogenic compounds is related to modulation of the ERK pathway, and in associated studies (see Supplementary Figure S2 at http://www.clinsci.org/cs/122/cs1220409add.htm) we found that the inhibition of Aβ production by lovastatin [a competitive inhibitor of HMG-CoA (3-hydroxy-3-methylglutaryl-CoA) reductase (a cholesterol-lowering drug)] was correlated with reduced ERK phosphorylation (whereas the γ-secretase inhibitor DAPT did not alter ERK signalling). Since ERK signalling is clearly dysregulated in AD [28,29], normalization of this pathway may still represent a therapeutic option (for PDMP-related compounds and other more specific ERK modulators), however, this does not appear to be the pathway by which these ceramide analogues inhibit Aβ production.

A recent study indicates that inhibition of the ERK pathway has no impact on cellular sAPPα secretion [53]. This finding is in agreement with the present study that shows inhibition of ERK activation by PDMP, PPMP and EtDO-P4, and their enantiomers, where applicable, does not significantly alter cellular sAPPα secretion. It is noteworthy, however, that some aspects of α-secretase activity may be modulated by ERK activation in a substrate-specific manner as ERK1 does control α-secretase-mediated processing of the cellular prion protein [54].

Previous research suggests that raft-associated GSLs promote amyloidogenic processing of APP via regulation of BACE and γ-secretase activities [9,11]. We therefore speculated that the inhibition of Aβ production by the ceramide analogue GCS inhibitors we used would be mirrored by siRNA silencing of GCS. Intriguingly, the ~75% reduction in cellular GSL levels that was induced by GCS knockdown had no impact on Aβ production. This may indicate that the D-enantiomers of the ceramide analogues we have used have additional actions (either direct or indirect) on APP processing that remain to be discovered and that inhibition of GSL synthesis alone (i.e. using siRNA targeting) may only be efficacious in situations where membrane GSL levels are increased above their physiological concentrations. Alternatively, it may be that modulation of very specific neutral or charged GSL species that are not highly expressed in the in vitro systems we have used is required to reveal an impact on Aβ production. Future studies would be required to examine the effect that the numerous individual GSL species may have on cellular APP processing. We acknowledge that specific GSLs play other roles (in addition to potentially regulating APP processing) that may impact on AD pathology. As a case in point, ganglioside G_M1 plays a crucial role in regulating the transition of monomeric Aβ to neurotoxic oligomers and the use of compounds such as D-EtDO-P4 to suppress this activity would represent a viable therapeutic target that is independent of its impact on APP processing.

It is noteworthy that D-threo-EtDO-P4 has been administered successfully in long-term studies in mice [21,55]. In addition, a potentially relevant GCS inhibitor {Eliglustat tartrate, (1R,2R)-octanoic acid [2-(2′,3′-dihydro-benzo [1,4] dioxin-6′-yl)-2-hydroxy-1-pyrrolidin-1-ylmethyl-ethyl]-amide-L-tartaric acid salt, formerly Genz-112638} is currently in human clinical trials to treat Type 1 Gaucher disease where it has been shown to reduce plasma GlcCer and G_M3 levels in association with improvement in a number of clinical manifestations of the disease [56–59]. The data from our group and others suggest that such GCS inhibitors may also be worth investigating as anti-amyloidogenic agents in relevant AD amyloidogenic mouse models and potentially in the human AD setting.

This work was supported by the Australian National Health and Medical Research Council [project grant number 568651]. B.G. is supported by a fellowship from the Australian Research Council (ARC Future Fellowship) [grant number FT0991986].

We are grateful to Professor Colin Masters and Dr Qiao-Xin Li for providing the W0-2 antibody, and to Michael Ho for maintaining the SH-SY5Y-APP cell line.

2AA

2-anthranilic acid

Aβ

amyloid β-peptide

Aβ40

Aβ-(1–40)

Aβ42

Aβ-(1–42)

AD

Alzheimer's disease

APP

amyloid precursor protein

BACE-1

β-site APP-cleaving enzyme

CHO cell

Chinese-hamster ovary cell

DAPT

N-[N-(3,5-difluorophenacetyl)-L-alanyl]-(S)-phenylglycine t-butyl ester

ERK

extracellular-signal-regulated kinase

EtDO-P4

D-threo-ethylenedioxy-1-phenyl-2-palmitoylamino-3-pyrrolidino-1-propanol

FBS

fetal bovine serum

GAPDH

glyceraldehyde-3-phosphate dehydrogenase

GCS

glucosylceramide synthase

GlcCer

glucosylceramide

GSL

glycosphingolipid

LacCer

lactosylceramide

MAPK

mitogen-activated protein kinase

MTT

3-(4,5-dimethylthiazol-2-yl)-2,5-2H-tetrazolium bromide

PDMP

D-threo-1-phenyl-2-decanoylamino-3-morpholino-1-propanol

pERK

phosphorylated ERK

PPMP

D-threo-1-phenyl-2-hexadecanoylamino-3-morpholino-1-propanol

PS

presenilin

sAPPα

soluble APPα fragment

siRNA

small interfering RNA

Hongyun Li conducted the experimental work with contributions from Ya Hui Hung and Brett Garner. Hongyun Li, Genevieve Evin, Andrew Hill, Ashley Bush and Brett Garner contributed to method development, establishment of cell lines, experimental design and data interpretation. Hongyun Li and Brett Garner wrote the paper with contributions from Genevieve Evin, Andrew Hill, Ya Hui Hung and Ashley Bush.