• ALS-parkinsonism dementia complex;
  • cycad;
  • glutamate release;
  • neurotoxicity;
  • NMDA receptor;
  • sterol β-d-glucoside


  1. Top of page
  2. Abstract
  3. Materials and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

The factors responsible for ALS-parkinsonism dementia complex (ALS-PDC), the unique neurological disorder of Guam, remain unresolved, but identification of causal factors could lead to clues for related neurodegenerative disorders elsewhere. Earlier studies focused on the consumption and toxicity of the seed of Cycas circinalis, a traditional staple of the indigenous diet, but found no convincing evidence for toxin-linked neurodegeneration. We have reassessed the issue in a series of in vitro bioassays designed to isolate non-water soluble compounds from washed cycad flour and have identified three sterol β-d-glucosides as potential neurotoxins. These compounds give depolarizing field potentials in cortical slices, induce alterations in the activity of specific protein kinases, and cause release of glutamate. They are also highly toxic, leading to release of lactate dehydrogenase (LDH). Theaglycone form, however, is non-toxic. NMDA receptor antagonists block the actions of the sterol glucosides, but do not compete for binding to the NMDA receptor. The most probable mechanism leading to cell death may involve glutamate neuro/excitotoxicity. Mice fed cycad seed flour containing the isolated sterol glucosides show behavioral and neuropathological outcomes, including increased TdT-mediated biotin–dUTP nick-end labelling (TUNEL) positivity in various CNS regions. Astrocytes in culture showed increased caspase-3 labeling after exposure to sterol glucosides. The present results support the hypothesis that cycad consumption may be an important factor in the etiology of ALS-PDC and further suggest that some sterol glucosides may be involved in other neurodegenerative disorders.

Abbreviations used

ALS-parkinsonism dementia complex


(S)-2-amino-3-(3-hydroxy-5-methylisoxazol-4-yl) propionic acid


β-methyl amino-alanine


S -(–)-β- N -oxalyl-α,β-diaminopropionic acid




campesterol/dihydrobrassicasterol β-d-glucoside


stigmasterol β-d-glucoside


β-sitosterol β-d-glucoside


d -(–)-2-amino-5-phosphonopentanoic acid


distilled H2O




ethyl acetate


lactate dehydrogenase

K–H buffer

Krebs–Heinseleit buffer






mass spectrometry


2,3-dioxo-6-nitro-1,2,3,4-tetrahydrobenzo [f] quinoxaline-7- sulfonamide




TdT-mediated biotin–dUTP nick-end labelling.

The constellation of neurological disorders on Guam – ALS-parkinsonism dementia complex (ALS-PDC) – has defied explanation for more than 50 years. A number of hypotheses have been proposed to explain the unusual features and incidences of these disorders (Andrews 1967; Kurland 1988). Many of these hypotheses were based on the notion of an environmental (neuro)toxic factor (Kurland et al. 1994). Early epidemiological reports focused on potential toxins associated with the consumption of flour made, by the indigenous Chamorro population, from the seed endosperm of a local variety of cycad palm (Cycas circinalis). Several compounds isolated from the cycad seeds, notably the amino sugar methylazoxyethanol β-d-glucoside (cycasin) and its aglycone, methylazoxymethanol (MAM), were found to be carcinogenic and to cause developmental abnormalities of the nervous system (Campbell et al. 1966), but were not shown to be toxic to neurons in adult animals in vivo. Several years ago, considerable renewed interest was generated by the identification of β-methyl amino-alanine (BMAA), an NMDA-receptor agonist present in unwashed cycad (Spencer et al. 1987). However, BMAA has been shown to be largely removed by the traditional practice in which the cut seeds are extensively washed before consumption (Duncan et al. 1990; U.-K. Craig, personal communication). The apparent absence of a neurotoxic factor combined with a decreasing overall incidence of ALS-PDC on Guam led to a general decline in interest of this disease (Kurland et al. 1994).

We have re-examined the ‘cycad’ hypothesis from the perspective that a potential neurotoxic factor would have to be present in the washed cycad product consumed in traditional Chamorro practice. Such a compound would have to be insoluble in water, would have to survive normal cooking temperatures, and would have to be sufficiently lipophilic to be able to cross the blood–brain barrier to exert its neurotoxic action.

To pursue these studies, we obtained washed cycad chips from Guam and used the extracts of these in a series of bioassays designed to examine neural excitability, second messenger pathways, and neurotoxicity. Using column chromatography, we isolated the putative active neurotoxic components of washed cycad. The toxicity and possible mechanisms of action of three sterol β-d-glucosides will be described in the present article.

Materials and methods

  1. Top of page
  2. Abstract
  3. Materials and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References


Cortical slices used in field potential recording, lactate dehydrogenase (LDH) release, glutamate release and protein kinase assays were made from sensory cortex of adult (> 60 days of age) male Sprague–Dawley colony rats maintained on a light : dark cycle (12 h : 12 h).

Single cell recording was made using fetal Sprague–Dawley rat cortex or hippocampal neuronal cultures or from adult hippocampal slices.

Adult (5–10 months of age) male CD-1 mice were fed washed cycad flour as a portion of their diet and subjected to a battery of behavioral tests (see Wilson et al. 2002).

Ethics of animal use

Animal experiments were performed under the auspices of the University of British Columbia Animal Care Committee (license no. A97-0150) and the Hospital for Sick Children Animal Care Committee of the University of Toronto. Both conform to the Canadian Council on Animal Care guidelines. Cell culture experiments were performed using embryonic human brain tissue as approved by the Ethics Committee of the University of British Columbia, Faculty of Medicine.


Buffer ingredients, β-sitosterol and stigmasterol, used to synthesize sterol β-d-glucosides, and LDH kits were all obtained from Sigma-Aldrich Canada Ltd. (Mississauga, Ontario, Canada). (S)-2-amino-3-(3-hydroxy-5-methylisoxazol-4-yl) propionic acid (AMPA), NMDA, d-(–)-2-amino-5-phosphonopentanoic acid (D-AP5, and 2,3-dioxo-6-nitro-1,2,3,4-tetrahydrobenzo [f] quinoxaline-7-) sulfonamide (NBQX), and 6,7-dinitroquinoxaline-2,3-dione (DNQX) were obtained from Precision Biochemicals Inc. (Vancouver, BC, Canada). 3-heptenoic acid, 2-amino-4-(phosphonomethyl)-(E)-(+, –) ([3H]CGP 39653) and [3H]glutamate were purchased from NEN/Mandel Scientific Co. (Guelph, Ontario, Canada). Dr G. E. Kisby of the Center for Research on Occupational and Environmental Toxicology, Oregon Health Sciences University, supplied methylazoxyethanol β-d-glucoside (cycasin) and methyl azoxymethanol (MAM). β-methyl aminoalanine (BMAA) was purchased from Sigma-Aldrich Chemicals; S-(-)-β-N-oxalyl-α,β-diaminopropionic acid (BOAA) was obtained from Research Biochemicals Inc. (Notick, MA, USA). The chemiluminescent ECL RPN-2209 (ECL) was obtained from Amersham Pharmacia Biotech (Bucks, UK). Other chemicals of analytical grade were from BDH Inc. (Vancouver, BC, Canada).

Cycad preparation and extraction

Cycad seeds of the species Cycas circinalis were obtained from Guam. Samples included uncut fresh ripe fruits containing seeds as well as seed sections (‘chips’), which had been washed between one and seven times using traditional methods (Duncan et al. 1990; U.-K. Craig, personal communication). The wash procedure has been previously demonstrated to remove almost all traces of BMAA (Duncan et al. 1990). In preliminary experiments, a 1 : 50 crude, water extract of each preparation was prepared and screened for biological activity using a cortical wedge preparation (Shaw et al. 1999) and by assays for LDH activity in a cortical slice as an index of cell death.

Applying such methods, different batches of seeds and washed cycad sections were screened for biological activity, as described below. Using the most biologically active batch of the processed cycad flour, 200 g of chips were immersed in 700 mL of methanol (MeOH) and subsequently extracted three times. The combined methanolic extracts were concentrated in vacuo and the resultant amorphous solid (2 g) was first bioassayed and then partitioned between ethyl acetate (EtOAc, 3 × 500 mL) and distilled water (dH2O, 1500 mL). The extracts from each partition were combined for total EtOAc or dH2O partitions and evaporated to dryness to yield 1200 mg of amorphous water-soluble material and 800 mg of organic material as clear oil. The extracts were bioassayed using LDH release and cortical wedge recording, as described below. The active EtOAc extract was chromatographed on Sephadex LH-20 in 4 : 1 MeOH/CH2Cl2 to give 524.5 mg of a late eluting fraction. This material was further fractionated using reverse-phase flash chromatography employing a step gradient from H2O to MeOH. A 101.8-mg fraction, eluting with 9 : 1 MeOH/H2O to MeOH, contained two bright purple staining spots when visualized on normal-phase thin-layer chromatography plates (eluted with 1 : 19 MeOH/CH2Cl2) with a standard vanillin/H2SO4 spray reagent. This fraction was chromatographed via normal-phase flash chromatography, employing a step gradient using CH2Cl2 followed by 1 : 19 MeOH/CH2Cl2. Four fractions were obtained: two were inactive; a mildly biologically active fraction (41.4 mg), eluted with CH2Cl2, contained a mixture of lipids and the free steroids, β-sitosterol, campestral or dihydrobrassicasterol, and stigmasterol; and a potently active fraction (31.2 mg), eluted with CH2Cl2, contained a mixture of lipids and the sterol glucosides β-sitosterol β-d-glucoside, campestral or dihydrobrassicasterol β-d-glucoside, and stigmasterol β-d-glucoside. Pure β-sitosterol β-d-glucoside (7.3 mg), a mixture of campesterol or dihydrobrassicasterol β-d-glucoside (3.65 mg), stigmasterol β-d-glucoside, and a fraction containing inactive lipid material was obtained from the latter active fraction via C18 reverse-phase HPLC using a Whatman Magnum-9 Partisil 10 ODS-3 column, with 19 : 1 MeOH/dH2O as eluent. Pure campesterol β-d-glucoside or dihydrobrassicasterol β-d-glucoside and stigmasterol β-d-glucoside were then obtained from reverse-phase HPLC using an analytical Alltech Econosil 5-µm column, with 39 : 11 MeCN/dH2O as eluent.

HPLC/mass spectroscopy

HPLC/mass spectrometry (MS) was performed at the Department of Pharmacological Sciences (University of British Columbia). A Micromass Quattro mass spectrometer and an HP 1090 Series 2 HPLC were employed to estimate the concentrations of toxins previously associated with cycad seeds. We compared cycad flour extracts to template standards for lipophilic moieties such as sterol glucosides and water-soluble compounds such as cycasin, MAM, BMAA, and BOAA. A 2-g quantity of cycad chips was crushed in 20 mL of dH2O, then sonicated to extract the cycasin, MAM, BMAA, and BOAA present. To extract lipophilic compounds, such as sterol β-d-glucoside, the sonicated flour was subsequently extracted in MeOH. The extracts were filtered using filter paper (Whatman no. 1, retention size ≈ 11 µm). The filtrates were dried and then redissolved in dH2O or MeOH, respectively (methylene chloride was also used to maximize the resuspension of extract) to give a ‘crude’ extract. The crude extract was diluted by 1 : 20 in a 50% MeOH, 50% dH2O solution (called the solvent phase). Fractions were passed through a Phenomenex Luna 00B-4256-B0 3-micron phenyl-hexyl, 50 × 2-mm column. We used the mixed solvent containing dH2O and MeOH to avoid formation of MeOH adducts. The solvent mix contained MeOH, dH2O, and formic acid, depending on the ability of the molecule to generate negative/positive ions and to dissolve readily. The ion source head was cleaned at the end of each experiment to ensure a 100% working ion generator. In this part of the experiment, the desired compound was exposed to the ion source to generate either negative or positive ions on the molecule. Applying 15–30 electron volts of collision-energy and 1.05 e−6 to 9.75 e−3 argon achieved fragmentation, which was performed to detect the daughter molecules produced from a designated peak after collision energy and argon had been applied.

Synthesis of sterol glucosides

Acetobromo-α-d-Glucose(2,3,4,6-Tetra-O-acetyl-α-d-Glucopyranoside Bromide) was added to the commercially purchased sterols. Silver triflate (CF3SO3Ag) was added followed by lutidine in CH2Cl2 (a 1 : 1 : 1 : 1 ratio was used for triflate : acetobromoglucose : lutidine : sterol). The resulting substance was synthetic sterol-α-d-glucoside (70–90% yield), which was not neurally active, as assessed by field potential recording and LDH-release assays (see below). When greater than fivefold triflate and acetobromoglucose was added, a more active substance (synthetic sterol-β-d-glucoside) (1–4% yield) was obtained along with acetylated sterol (70–90% yield).

Solubility of sterol glucosides

The lipophilic nature of the sterol glucosides required the use of a solvent that could dissolve these compounds and yet be non-toxic to cells. We found pyridine to be the best candidate as it completely dissolved the sterol glucosides and was not toxic up to 0.01%, as confirmed in our LDH release and cortical field potential recording experiments. Other solvents such as Tween-20 and dimethyl sulfoxide (DMSO) were found to be more toxic at concentrations promoting solubility of the sterols. In some instances, precipitation would form when sterol glucosides already dissolved in pyridine were added to Krebs-Heinseleit (K–H) buffer. In these cases, we sonicated the samples to ensure complete homogeneity of test solutions.

LDH-release measurements

Cortical slices were prepared from adult (> 60 days old) male rat (Sprague–Dawley) neocortex, as described previously (see Shaw et al. 1999) and placed in tissue culture wells containing 0.5 mL of K–H buffer supplemented with 0.0004% H2O2 and 1 mg/mL glucose. Extensive previous studies have demonstrated that this and similar media support cellular activity for prolonged time-periods (see Shaw et al. 1999). Hydrogen peroxide, added as the source of molecular oxygen, was not deleterious at this concentration (Shaw et al. 1999). [In our preliminary experiments, hydrogen peroxide did not affect LDH release at a concentration of up to 1 mm (i.e. 0.0034%; data not shown)]. All slices were washed twice with K–H buffer containing 1.2 mm Mg2+ for 20 min at room temperature (≈ 25°C) before incubation (for 1 h at 37°C) in K–H buffer containing the test compounds. Test compounds included the isolated cycad extracts of β-sitosterol β-d-glucoside, synthetic β-sitosterol β-d-glucoside, and NMDA. Several of these were tested alone or in combination with a glutamate receptor antagonist, D-AP5, and compared to control slices maintained in buffer alone. For additional comparison and to establish the limits of the method, freeze–thawed slices were used as a positive control. Alternatively, some slices were incubated in buffer containing 1.2 mm Mg2+ in order to diminish spontaneous neural activity (see Shaw et al. 1999). At the end of the 1-h incubation period, three samples (100 µL of buffer from each sample) were taken from each well. LDH assays were performed on these samples using an LDH diagnostic kit (Sigma) following the manufacturer's protocol with some modifications. In brief, 0.5 mL of pyruvate solution was mixed with 0.5 mg of preweighed β-NADH. A 100-µL volume of slice medium (K–H; free of slices) was added to the mixture and incubated for 30 min at 37°C. A 0.5-mL volume of Sigma coloring reagent (2,4-dinitrophenylhydrazine in HCl, 2 mg/mL) was added for color development and the mixture was incubated for 20 min at room temperature. A 5-mL volume of 0.4 N NaOH was added to each tube to stop the enzyme reaction and, after 5 min, the optical density was read at 440 nm. Standard curves were prepared for each assay using different concentrations of pyruvate solution (0–960 Units). LDH activity (in International Units) was calculated from the standard curve and normalized by total protein content of each slice as determined by using a modified Lowry protein assay (Peterson 1979). One International Unit represents the amount of enzyme required to convert 1 µmol of substrate to product/min at room temperature.

Cortical field potential recording from the wedge preparation

Cortical ‘wedges’ were prepared as described previously (Shaw et al. 1999). In brief, adult male rats were anesthetized with CO2, decapitated, and a cortical block rapidly removed and placed in cold K–H buffer containing: 124 mm NaCl, 3.3 mm KCl, 25 mm NaHCO3, 10 mm glucose, 1.2 mm KH2PO4, 2.4 mm CaCl2, and 1.2 mm MgCl2, bubbled with 5% CO2/95%O2, with a final pH of 7.4. The cortical block was sectioned into 500-µm thick slices using a Vibratome (Campden, Leicestershire, UK) and the slices cut into pie-shaped wedges in which the white matter formed the narrow edge of the wedge. Each wedge was placed on a net across a grease gap between two fluid-filled chambers. The cortical side of the wedge was bathed in room temperature buffer without Mg2+ the callosal portion was bathed in buffer containing Mg2+ to minimize neural activity. Field potentials were differentially recorded between the two chambers using two Ag/AgCl electrodes. Recordings from up to six wedges, each in individual chambers, could be made simultaneously for each experiment. The wedges were continuously perfused on the cortical side with oxygenated, Mg2+-free buffer using a gravity-feed system. Using this system, drugs could be rapidly substituted for control media to examine response characteristics. Responses were recorded on LabView after amplification and A/D conversion and the traces were charted in Excel for Windows. Wedges typically survived for up to 8 h. Statistical analysis of peak response amplitude was performed by one-way anova (two-tailed) using Bonferroni's post-hoc test with GraphPad Prism.

Single cell recording

Electrophysiological recordings were made as described previously (Wang and Salter 1994). In brief, primary cultured hippocampal neurons were prepared from fetal rat (Sprague–Dawley) hippocampus. Subsequently, a whole-cell patch-clamp technique was used in 8-day- or 3-week-old cultures with sterol glucosides applied by puffing (campesterol or dihydrobrassicasterol β-d-glucoside, 100 µm). Whole-cell recordings were made using an Axopatch-1D amplifier (Axon Instruments, Foster City, CA, USA) at a holding potential of −60 mV. A recording chamber was perfused at room temperature with extracellular solution containing: 140 mm NaCl, 5.4 mm KCl, 1.5 mm CaCl2, 10 mm HEPES, 25 mm glucose and 0.001 mm tetrodotoxin (pH 7.4, 310–320 mOsm) and the recording pipettes (4–5 mΩ) were filled with intracellular solution containing: 135 mm CsCl, 2 mm MgCl2, 0.1 mm CaCl2, 10 mm EGTA, 10 mm HEPES and 4 mm K-ATP (pH 7.2, 300–310 mOsm). Both NMDA (50 µm) and sterol β-d-glucoside (up to 500 µm) were applied (100 ms) by pressure through a drug pipette closed to the recording cell. In a few experiments, single cell recordings were made from adult hippocampal slices.

NMDA receptor-binding studies

Cortical synaptic membrane fractions were prepared as described previously (Shaw et al. 1999). Briefly, membrane aliquots (final volume 500 µL containing 200–300 µg of protein) were incubated with 2 nm[3H]CGP 39653 in 50 mm Tris acetate for 120 min at 4°C, and rinsed by filtration through Whatman GF/B filter disks with 2 × 5-mL volumes of cold buffer. Filters were placed into scintillation vials containing NEN Formula 989 for a minimum of 12 h before being counted in a Beckman LS6000 scintillation counter. Competition assays were run with NMDA, glutamate, or a synthesized β-sitosterol β-d-glucoside, over a concentration range of 10−3m to 10−9m. Curves were best fit in all cases by a single-site model using non-linear, least-squares regression analysis (GraphPad Prism).

[3H]Glutamate release studies

Cortical brain slices were prepared as described above. Slices were rinsed twice for 5 min in Mg2+-containing K–H buffer, pH 7.4. Then, 100 µm cold glutamate, 20 µm D-AP5 (NMDA antagonist) and 10 µm DNQX (AMPA antagonist) were added in 20 mL of K–H buffer. [3H]Glutamate (10 nm) was added to the mixture which was then incubated for 1 h at 37°C in an oxygenated atmosphere (O2/CO2 = 95/5%). Cycad and other treatments were performed in 500 µL of Mg+2-free buffer in tissue culture wells using different concentrations and different time intervals of incubation. The supernatant was removed at the end of the incubation time, placed in vials containing NEN Formula 989, and the radioactivity counted ina Beckman LS6000 scintillation counter. Results were normalized to the dpm counts of control slice supernatant.

Human astrocytes in culture

Four series of primary culture were prepared from embryonic human brains of 12–15 weeks of gestation (Nagai et al. 2001; Satoh and Kim 1994). Brain tissues were dissected into small blocks and incubated in phosphate-buffered saline (PBS) containing 0.25% trypsin and 40 µg/mL DNAse I for 30 min at 37°C. Dissociated cells were suspended in Dulbecco's modified Eagle medium (DMEM) supplemented with 5% horse serum, 5% fetal bovine serum, 5 mg/mL glucose, 20 µg/mL gentamicin and 2.5 µg/mL amphotericin B (feeding medium). The cells were plated at a density of 106 cells/mL in T75 culture flasks and incubated at 37°C in an incubator with a 5% CO2/95% air atmosphere. After 2–4 weeks, cells grown in the flasks consisted of a confluent basal layer of flat astrocytes, free-floating microglia, and small, round or ovoid neurons on top of the astrocyte layer. Microglia floating in the medium were then removed; neurons and a small number of oligodendrocytes were also removed by vigorous shaking, and astrocyte-enriched cell preparations were obtained from plastic surface-adherent cells by a brief incubation in PBS containing 0.1% trypsin and 1 mm EDTA. To confirm the purity of cell preparation, cell type-enriched cultures were processed for immunolabelling with cell type-specific markers, β-tubulin isotype III for neurons, glial fibrillary acidic protein (GFAP) for astrocytes, Ricinus communis agglutinin-1 (RCA-1) for microglia and galactocerebroside (GalC) for oligodendrocytes. Mouse anti-β-tubulin isotype III monoclonal antibody, rabbit anti-GFAP antibody and biotinylated RCA-1 were obtained from Sigma. Mouse anti-GalC monoclonal antibody was obtained from a hybridoma cell line grown in our laboratory (Ranscht et al. 1982). Purity of the astrocyte population was > 95%, as evaluated by immunocytochemistry performed according to the methods described previously (Satoh et al. 1996).

Immunoblotting and protein kinases

Cortical tissues were prepared by the method described in Shaw et al. (1999). Slices were placed in K–H buffer containing different compounds for 15 min. These compounds included NMDA and β-sitosterol β-d-glucoside, with and without D-AP5. Tissue samples were homogenized in sodium dodecyl sulphate (SDS) sample buffer (containing 75 mmβ-d-glycerophosphate and 1 mm sodium orthovanadate, pH 7.2). The homogenate was then centrifuged for 30 min at ≈ 100 000 g to remove impurities, including lipids. SDS–polyacrylamide gel electrophoresis (SDS–PAGE) was performed to separate the homogenate proteins (based on size) and the resolved proteins were subsequently transferred electrophoretically onto a nitrocellulose or polyvinylidenedifluoride (PVDF) membrane (Baraban et al. 1993). Membranes were blocked by Tris-buffered saline (TBS) buffer containing 1.5% bovine serum albumin (fraction V) and 2.5% skimmed milk, and then washed with TBS containing 0.05% Tween-20 (TTBS). Membranes were then probed with commercial primary and secondary antibodies against various kinases, following the method of Baraban et al. (1993). ECL was used to illuminate positive bands and the membranes were exposed to Kodak BioMax MR Film. The film was developed and band detection performed to measure dye front (DF) for band size and band shifting owing to activation or inhibition of specific kinases. Densitometry was performed using Power-Mac to quantify the results (band shifts).Kinases studied included PKC-β (78 kDa), CDK-2 (34 kDa), CaMKII (58 kDa), Cot (55 kDa), Pim-1 (37 kDa), PKG-1 (75 kDa), S6K (65 kDa), PKA (41 kDa), PKB-2 (59 kDa), ErK-1 (42 kDa), ErK-5 (85 kDa), RsK-1 (90 kDa), and MKK-4 (45 kDa).

In vivo studies on cycad-fed mice

For details see Wilson et al. (2002).


Data were analyzed for significance by one-way anova (two tailed) using Dunnett's and Bonferroni's for post-hoc tests.


  1. Top of page
  2. Abstract
  3. Materials and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

Preliminary data with crude cycad extract: organic fraction

In preliminary experiments, we found that both the fresh cycad seed extract and the product washed seven times (7× washed) gave depolarizing responses (Fig. 1a), which could be selectively blocked by the NMDA receptor antagonist D-AP5 and the non-competitive antagonist MK-801. In contrast, the AMPA receptor antagonist DNQX had no consistent effect on cycad-induced responses. Similarly, 7× washed extracts led to rapid cell death, as determined by LDH release in a cortical slice, both of which effects could be blocked by the addition of D-AP5 (Fig. 1b). Seven times washed cycad extract often gave electrophysiological responses and LDH activity comparable to that of whole cycad extract, indicating that a toxic factor in this batch of cycad seeds was not removed by processing in the traditional manner.


Figure 1. Cortical electrophysiological response and LDH release to crude cycad extract. (a) Cortical ‘wedge’ recording of adult rat neocortex. Pre-incubation with MK801 (10 µ m ) or D-AP5 (20 µ m , not shown) blocked the NMDA (50 µ m ) and cycad-induced responses of 1 : 50 or 1 : 100 dilutions; NBQX (10 µ m ) blocked the AMPA (20 µ m ) response, but had no effect on the cycad response (data not shown). In each trace in this and subsequent figures, downward arrows indicate the onset of drugs application and each such application was for 2 min. Similar results were obtained from a minimum of five other cortical wedges in this experiment. The experiment was repeated with multiple wedges three times. (b) LDH release assay as a measurement of cycad toxicity in rat cortical slices. Cortical slices were assayed for LDH release following exposure to various compounds for1 h at 37°C. Crude cycad in the same concentration as in (a) gave greater LDH release than evoked by NMDA. D-AP5 attenuated cycad-induced LDH release. This figure represents triplicate measurements. The experiment was repeated three times. * p  < 0.05 , one-way anova (two-tailed) compared with control; # p  < 0.05 in the presence of NMDA receptor antagonist D-AP5. Error bars indicate SEM.

Download figure to PowerPoint

HPLC/MS: isolation of cycad neurotoxins

Based on the above studies, we proceeded to isolate increasingly specific cycad fractions, focusing our efforts on one batch of the washed cycad product.

Three separate batches of cycad chips (7× washed May 1998; 7× washed September 2000; and 5× washed May 1998), collected from various sites on Guam, are reported. The May 1998 batch (7× washed) was exclusively used in the experiments described in this article. Values for the 7× washed September 2000 and 5× washed May 1998 batches are shown for comparison only.

The extracts were subdivided (as described above in the Materials and methods) into aqueous and organic fractions and screened for potency using cortical field potential recording and LDH-release assays, as described above. HPLC/MS was performed and our washed cycad flour samples were found to contain no traces of cycasin or MAM – compounds previously linked to ALS-PDC (Kisby et al. 1992). Neither did our samples contain high concentrations of BOAA (the present data detected 1.92e−2, or 0.14 µg/g) – reported to cause lathyrism in animals and humans (Spencer et al. 1987) (see Table 1). BMAA concentrations were significantly lower than those reported by Kisby et al. (1992) (8.00e−5−3.00e−3µg/g vs. up to 18.4 µg/g). The compounds found to be toxic, identified below as sterol β-d-glucosides, were present at concentrations of ≈ 21–155 µg/g.

Table 1.  HPLC/mass spectrometry (MS) analysis of three separate samples of washed cycad flour
Compound7×, May 19987×, September 20005×, May 1998Kisby et al. (1992)aDuncan et al. (1990)b
  1. ND, not detected and the detection limit for total comparison was 8.00e−5µg/g. Washed values are per 10 µL of a 1 : 20 dilution of 20 g of flour in 20 mL of MeOH. Two separate batches of cycad chips, collected from various sites on Guam, are reported (May 1998; September 2000). The May 1998 batch (7×washed) was exclusively used in the experiments described in this article. Values for the 7×washed September 2000 and 5×washed May 1998 batches are shown for comparison only. aData from Kisby et al. (1992) used cycad flour of unknown wash times. bDuncan et al. (1990) report values for a 1×washed cycad. The concentrations of BMAA, cycasin, MAM, and BOAA detected in this analysis were significantly lower than that reported by Kisby et al. (1992). Sterol β-d-glucoside fractions were present at concentrations between 21 and 155 µg/g. This experiment was performed in duplicate.

CycasinND0.12 µg/g5.20e−2 µg/g4.00e−3 to 75.9 µg/g
MAMND0.61 µg/g0.52 µg/g0.23–3.74 µg/g
BOAA0.14 µg/gND1.92e−2 µg/g
BMAA3.00e−3 µg/g1.20e−4 µg/g8.00e−5 µg/gND to 18.4 µg/gND to 1.26e2
D-211.5 µg/g84.5 µg/g47.4 µg/g
D-1-29.96 µg/g71.1 µg/g20.6 µg/g

Identification of compounds

The active compounds were identified as the known sterol glucosides, β-sitosterol β-d-glucoside (D-2), stigmasterol β-d-glucoside (D-1-2), and campesterol/dihydrobrassicasterol β-d-glucoside (D-1-1) (see Fig. 2), by detailed analysis of NMR and MS data obtained for the individual pure compounds. It was not possible to determine if, in the latter case, the one active compound was campesterol β-d-glucoside or dihydrobrassicasterol β-d-glucoside on the basis of the NMR and MS data. In all subsequent data and figures, we will refer to these compounds as D-2, D-1-1, and D-1-2.


Figure 2. Isolation procedure for potential cycad neurotoxins: sterol glucosides. A 200-g sample of cycad flour was extracted in 1500 mL of methanol. The supernatant was dried and further resuspended and partitioned between ethyl acetate and dH 2 O. Serial purification gave final concentrations (per total initial sample) of 7.3 mg of β-sitosterol β- d -glucoside (D-2), 0.1 mg of campesterol/dihydrobrassicasterol β- d -glucoside (D-1-1), and 0.3 mg of stigmasterol β- d -glucoside (D-1-2).

Download figure to PowerPoint

Bioassays using isolated cycad fractions

Field potential recordings for each of the isolated steroidal glucosides showed a rapid depolarization that could be blocked with NMDA antagonists. In Fig. 3(a) we show the cortical field potential responses of D-2 compared to other known sterol glucosides, ouabain and emicymarin, versus the response to NMDA. Both the cycad compound and NMDA gave a depolarizing response that could be blocked by D-AP5. Ouabain, emicymarin and the aglycone β-sitosterol gave hyperpolarizing responses. Figure 3(b) shows the cortical field potential responses to sonicated D-2 and D-1-1 compared to NMDA. Sonicated sterol glucoside fractions generated significantly larger responses than the non-sonicated compounds and all were blocked by D-AP5 (not shown). LDH assays using the sterol β-d-glucosides showed toxicity that was blocked by D-AP5 (Figs 3c and d). Initial estimates with limited samples showed that D-2 appeared to show equivalent toxicity to NMDA, while D-1-1 was more potent in single-concentration assays. Owing to a limited number of samples, D-1-2 was not tested for LDH release.


Figure 3. Bioassays using isolated sterol glucoside fractions. (a) Cortical field potential response to β-sitosterol β- d -glucoside (D-2). The D-2 (15 µ m ) and NMDA (50 µ m ) both gave depolarizing responses; preincubation with the NMDA receptor agonist D-AP5 (20 µ m ) blocked both responses. β-sitosterol (β-S), the aglycone form of D-2, did not depolarize cells. (b) Sonicated D-2 (50 µ m ) and β-campesterol/ dihydrobrassicasterol β- d -glucoside (D-1-1) 20 µ m gave large depolarizations that were attenuated by D-AP5. (c) LDH release in cortical slices treated with D-2 (50 µ m ) gave equivalent values to those evoked by NMDA (50 µ m ); the aglycone form of β-S (50 µ m ) did not give LDH release. The effects of NMDA and D-2 were both attenuated by D-AP5 (20 µ m ). (d) LDH release in cortical slices treated with β-campesterol or dihydrobrassicasterol β- d -glucoside (25 µ m ) gave equivalent LDH release to that evoked by NMDA (50 µ m ). The effects of NMDA and β-campesterol or dihydrobrassicasterol β- d -glucoside were attenuated by D-AP5 (20 µ m ). These data represent triplicate measurements and were repeated three times. * p  < 0.05; ** p  < 0.01 One-way anova (two-tailed) compared with control; # p  < 0.05; ## p  < 0.01 in the presence of NMDA receptor antagonist D-AP5. Error bars indicate SEM.

Download figure to PowerPoint

Role of NMDA receptor activation

NMDA antagonists were able to attenuate/block the effects of the various sterol fractions, leading to speculation that these exerted some of their action via NMDA receptor channel activation. To test this hypothesis, we began a series of studies employing whole-cell recordings from single neurons in vitro as well as competition studies to test for binding to the NMDA receptor.

Single cell recording

We examined the action of the isolated sterol β-d-glucosides on three single-cell preparations as described above in the Materials and methods. Using a whole-cell patch-clamp technique in 8-day-old-cultured rat hippocampal neurons, puffing of D-1-1 (100 µm) onto the cell did not invoke any response. Similar results were obtained using 3-week-old-cultured neurons (Fig. 4a). There was also no response in hippocampal CA1 pyramidal neurons to D-2 (500 µm) when applied to adult (≈ 60 days of age) rat brain slices. In contrast, NMDA (50 µm) applied using the same methods evoked a typical NMDA receptor-mediated response (Wang et al. 1994) under all of these conditions.


Figure 4. Test of direct action of β-sitosterol β- d -glucoside (D-2) on NMDA receptors. (a) Hippocampal single cell recording. The evoked whole-cell current was recorded using cultured hippocampal neurons. NMDA (50 µ m ), used as a control, induced a typical inward-going current; D-2 (500 µ m ) gave no response. Bars above traces show application of compounds. (b) Competition binding assay of NMDA and synthetic β-sitosterol β- d -glucoside (β-SG). The binding data shows that β-SG does not compete with [ 3 H]CGP for binding to the NMDA receptor. The binding was measured in triplicate in each of the three separate experiments, which are combined in this graph. Error bars indicate SEM.

Download figure to PowerPoint

Competition binding assays

Competition assays for the NMDA receptor antagonist [3H]CPG 39653 in adult rat cortical synaptosomal membranes were made comparing a synthesized version (see below) of sterol β-d-glucoside to NMDA or glutamate. Both glutamate and NMDA competed for binding; the sterol β-d-glucoside did not (Fig. 4b).

Mechanism of action of sterol glucosides

[3H]Glutamate release experiments

Given the results of the preceding experiments, we performed experiments to determine whether synthesized sterol glucosides were able to induce a glutamate release that could account for the observed blocking action of D-AP5. To test this hypothesis, we preloaded adult rat cortical slices with [3H]glutamate before exposure to D-2, D-1-2 or NMDA for 30 min at 37°C. Figure 5 demonstrates that exposure of cortical slices to NMDA or to the various sterol β-d-glucosides led to a significant release of glutamate that was calcium dependent and blocked by the co-incubation with D-AP5.


Figure 5. Glutamate released from cortical cells by sterol glucosides. (a) β-sitosterol β- d -glucoside (D-2) application at 25 µ m in a cortical slice preparation induced glutamate release similar to that induced by NMDA (50 µ m ). (b) Similar results were obtained using stigmasterol β- d -glucoside (D-1-2) 25 µ m . In (b) the dependence of Ca 2+ on glutamate release following NMDA or sterol glucoside was examined under conditions of 0 or high (2 m m ) Ca 2+ . (c) Human astrocytes were treated with D-2 (25 µ m ), with either 0 or 2 m m Ca 2+ in the buffering media. Glutamate release induced by D-2 was larger than that evoked by 150 m m KCl (0 Ca 2+ ). These experiments were performed in triplicate and repeated twice in each tested sterol glucoside. * p  < 0.05; ** p  < 0.01 One-way anova (two-tailed) compared with control; # p  < 0.05; ## p  < 0.01 in the presence of NMDA receptor antagonist D-AP5. Error bars indicate SEM.

Download figure to PowerPoint

Effects on protein kinase activity

Glutamate receptor activation and regulation is known to be highly dependent on protein kinase activity (for references see Pasqualotto and Shaw 1996). To further probe similarities to glutamatergic mechanisms, we examined the ability of D-1-1 and D-2 to influence protein kinase activity. Adult rat cortical slices were exposed to NMDA, D-1-1, or D-2 for 15 min at 37°C. Following exposure, slices were homogenized and the protein prepared for Western blot labeling with protein kinase ‘cocktails’ consisting of the test antibodies. The outcome of these experiments is shown in Fig. 6. Both NMDA and sterol glucoside fractions gave qualitatively similar changes in the level of expression of CDK-2 and PKC-β. The effects of both NMDA and sterol glucosides could be attenuated by D-AP5. The expression level of other protein kinases tested (including ErK-1, RsK-1, Cot and PKB-2) did not show a major change under the same conditions of stimulation.


Figure 6. Western blots and quantitation of cyclin-dependant kinase-2 (CDK-2) and protein kinase C-β (PKC-β) in an in vitro rat neocortical slices exposed to NMDA or isolated cycad fractions. (a) Rat cortical slices were treated with isolated campesterol/dihydrobrassicasterol β- d -glucoside (D-1-1) solublized (15 µ m ) in pyridine. The upward band shifts in CDK-2 indicates that both D-1-1 and NMDA (50 µ m ) activate this kinase. The effects of both were blocked by co-incubation of the sections with NMDA receptor antagonist D-AP5 (20 µ m ). Anisomycin (Aniso) (dissolved in DMSO), which induces apoptosis, did not activate CDK-2. (b) Quantification for the CDK-2 band shift by densitometry. (c) Band shift in the PKC-β indicates an up-regulation of the kinase to NMDA (50 µ m ) and D-2 (15 µ m ); both NMDA and D-2 were attenuated by D-AP5. (d) Quantification for the PKC-β band shift. The experiments were performed in triplicate; (b) and (d) * p  < 0.05; ** p  < 0.01 One-way anova (two-tailed) compared to control; # p  < 0.05 in the presence of NMDA receptor antagonist D-AP5. Error bars indicate SEM.

Download figure to PowerPoint

Studies of synthetic β-sitosterol β-d-glucoside

In most of these experiments, limited quantities of the isolated sterol glucoside precluded dose–response and time-course experiments. Largely owing to the relative difficulties associated with isolating large quantities of β-sitosterol-β-d-glucoside (β-SG) from cycad for further experimentation, we synthesized a β-SG that was identical to the natural compound. Figure 7 shows field potential recording and LDH release from rat cortical slices exposed to the synthetic compound, showing that its action closely resembles that of the natural compound. Like the former, the synthetic β-SG gave a depolarizing field potential (Fig. 7a) that was highly toxic, as assessed by LDH release (Fig. 7b, 7c). A synthetic cholesterol-β-d-glucoside (β-CG) was also prepared that was able to induce a depolarizing field potential (Fig. 7a) and was toxic, as measured by LDH release (Fig. 7c). The field potential recording and LDH release caused by these two compounds were blocked by D-AP5.


Figure 7. Electrophysiological field potential recording and LDH release using synthetic β-sitosterol β- d -glucoside (β-SG). (a) Field potential responses in a cortical wedge preparation were compared to NMDA (20 µ m ), β-SG (50 µ m ), and cholesterol β- d -glucoside (β-CG) (50 µ m ). Each of these compounds gave depolarizing responses; preincubation with NMDA receptor antagonist D-AP5 (20 µ m ) mostly blocked the depolarizations. (b) LDH release in cortical slices treated with β-SG or β-CG. β-SG (50 µ m ) gave an equivalent LDH release to that evoked by NMDA (50 µ m ). The effects of NMDA and β-SG were attenuated by D-AP5 (20 µ m ). (c) LDH release on cortical slices treated with β-SG or β-CG. The β-CG (50 µ m ) gave equivalent LDH release to that evoked by NMDA (50 µ m ). The effects of NMDA and β-CG were blocked by D-AP5 (20 µ m ). These experiments were performed in triplicate and repeated at least five times. * p  < 0.05; ** p  < 0.01 One-way anova (two-tailed) compared with control; # p  < 0.05 in the presence of NMDA receptor antagonist. Error bars indicate SEM.

Download figure to PowerPoint

Synthetic β-SG applied to astrocytes led to a 400% increase in Caspase-3 expression after 30 min exposure following incubation for 24 h at 37°C. Similar results were found for NMDA (+ 350%) and β-CG (+ 750%). These data are illustrated in Fig. 8.


Figure 8. Immunohistochemical assessment of cycad toxicity in human fetal astrocytes using synthetic β-SG. Caspase-3 activation in human astrocytes following exposure to β-SG (100 µ m ) compared with NMDA (100 µ m ), β-CG (100 µ m ), and hydrogen peroxide (H 2O2) 175 µm. Control: DMEM. These experiments were performed in triplicate and repeated at least three times. *p < 0.05; **p < 0.01; ***p < 0.001. One-way anova (two-tailed) compared with control. Error bars indicate SEM.

Download figure to PowerPoint

In vivo studies

A detailed series of behavioral and histological measurements were carried out on CD-1 mice fed washed cycad flour as a portion of their daily diet. These data are reported in detail in Wilson et al. (2002). The cycad flour used contained various sterol glucoside fractions, but only trace amounts of previously described cycad toxins (see Table 1; 7x washed cycad, May 1998 batch). Treated mice developed behavioral outcomes affecting motor and cognitive functions and also showed neurodegeneration of some regions of the CNS. Examples of the effect in cerebral cortex are shown in Fig. 9.


Figure 9. Histological assessment of neurodegeneration in mice fed cycad flour containing sterol glucosides ( Wilson et al. 2002 ). (a) TUNEL labeling in cortex of a cycad-fed mouse. (b) TUNEL labeling in equivalent cortical section of a control mouse. (c) TUNEL-positive patches of cells in cycad-fed mouse cortex. Calibration bar: (a) and (b) 25 µm; (c) 10 µm. (d) Cortical thickness measurements (µm) (from Nissl stained sections). Open bars: control mice; black bars: cycad-fed mice. V1, primary visual cortex; LEnt, lateral entorhinal cortex; M1, primary motor cortex; S1 primary somatosensory cortex; and Pir, piriform cortex. N = 10–18 each area. * p  < 0.05; ** p  < 0.01 One-way anova (two-tailed) compared with control; Error bars indicate SEM.

Download figure to PowerPoint


  1. Top of page
  2. Abstract
  3. Materials and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

The present results demonstrate that we have isolated and identified novel neurotoxins present in washed cycad seed. A search of the literature revealed that similar compounds have been previously suggested to exert neurotoxic effects (Ambike and Rao 1967). These compounds –β-sitosterol β-d-glucoside, campesterol or dihydrobrassicasterol β-d-glucoside, and stigmasterol β-d-glucoside – appear to have distinct effects on NMDA receptor activation, as NMDA antagonists can block their various actions in vitro. These actions include induction of depolarizing field potentials resembling those of NMDA, activation/inhibition of some of the same protein kinases as NMDA, the release of glutamate, and LDH release. While these results are all consistent with sterol β-d-glucoside action through NMDA receptors, this interpretation is seemingly at odds with the failure to find a direct activation of neurons in single cell recording experiments or competition binding for NMDA receptors (Figs 4a and b). However, a number of factors might serve to explain the apparent discrepancy. First, most single cell work was performed on neurons grown in culture from animals killed early in post-natal life. As age is a critical variable in the expression of various NMDA receptor subtypes (Monyer et al. 1994), the absence of a cellular response to the sterol glucoside indicates that these compounds may bind selectively to NMDA receptor subunits expressed only later in development. Another possibility is that the cell type affected in adult cortical slices is not the same as that in the culture. A further possibility is that the means of application in the single cell preparation (i.e. puffing) versus the bath application in the adult cortical wedge experiments may not expose NMDA receptors to sufficient concentrations of compound to promote channel opening. The results of the glutamate-release studies (Fig. 6), however, demonstrate that sterol β-d-glucoside application can lead to a large amount of glutamate released. It is thus highly probable that the effects we observed are caused by the indirect actions of glutamate released in the slice acting ultimately via NMDA receptor activation. As large-scale glutamate release would not be expected to occur in close proximity to the recording electrode in culture, this may account for the present results. Finally, we consider the possibility that the interactions with the NMDA receptor occur as has been recently demonstrated for another sterol, estrogen, which induces the phosphorylation of a tyrosine residue on NMDA receptors via a src tyrosine kinase/mitogen-activated protein (MAP) kinase pathway (Bi et al. 2000).

Regardless of the exact mode of action, the present results demonstrate that β-sitosterol β-d-glucoside, campesterol/dihydrobrassicasterol β-d-glucoside, and stigmasterol β-d-glucoside are acutely toxic to cells in adult rat cortex. A synthetic version of the former is toxic to cells in adult cerebral cortex slices and to fetal astrocytes in culture. Such results are in keeping with the toxic actions of other compounds containing sterol backbones, notably ouabain (Cargnelli et al. 1994) and digitalin (Piccone et al. 1995), both able to interact with membrane proteins. As such, the three sterol β-d-glucosides we have isolated join these compounds as being neurally active by virtue of particular molecular characteristics. Some of the latter can now be described as follows. The β-sitosterol aglycone is non-toxic, indicating that the sterol itself is not the active component in membrane interactions. This interpretation is supported by our preliminary studies with synthetic sterol β-d-glucosides, which are also toxic, but not in the aglycone form. In fact, the sterols themselves (like estrogen) could be neuroprotective over some dosage ranges (Bi et al. 2000).

One crucial difference between these three sterol β-d-glucosides and other similar sterol glucosides, such as ouabain, may reside in the way that their side-chains (e.g. attached to C20 position) are constructed. The other important difference may be a result of the number of double bonds in the side-chains and within the sterol moiety, the chiral orientation of the glucose-sterol bond (e.g. α or β), and the number of glucose molecules attached to the polar head (C3 position) of the steroid backbone. In regard to glucose number, Masayuki and Yasuo 1994) have shown that for various sterol glucosides toxicity decreases by any deviation of the number of sugar molecules from one (e.g. more than or less than one).

Parametric experiments are ongoing in an attempt to determine whether sterol β-d-glucosides administered in vivo are able to cross the blood–brain barrier to affect CNS neurons. Our current data in mice fed washed cycad flour suggests that some component of the flour, presumed to be the sterol β-d-glucosides, leads to cognitive and motor dysfunction and loss of neurons in various CNS fields (Fig. 9 and Wilson et al. 2002). In addition, in future experiments we hope to quantify the toxic actions of these compounds as functions of dose and duration of exposure, animal age, areas/cell types of CNS affected, and further clarify both the mechanisms of binding of these sterol glucosides to neural cells and those leading to cell death. An important future goal will be to determine if the sterol β-d-glucoside-induced neuronal damage is progressive once toxin exposure ceases; the latter would be a predicted feature of a toxin presuming to model human neurological disorders (e.g. ALS-PDC). Another goal will be to determine the conditions under which such sterol glucosides can be either synthesized or degraded in vivo, particularly in humans. In relation to the occurrence of ALS-PDC, it is interesting to speculate that abnormalities in either case could be related to neural degeneration if the sterol aglycone is glycosylated, or alternatively, by a failure to hydrolyze and detoxify the sterol β-d-glucoside.

Although much work remains to be carried out, the present data may lead to a re-evaluation of the hypothesis that cycad consumption is linked to ALS-PDC on Guam and suggests that toxins related by chemistry or action may play a role in other neurological disorders.


  1. Top of page
  2. Abstract
  3. Materials and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

Grant supported for these experiment was provided by ALS Association and the Scottish Rite Charitable Foundation of Canada (to CAS), National Science and Engineering Research Council (NSERC) (to RJA), Canadian Institutes of Health Research (CIHR) (to YTW), Kinexus Bioinformatics Corporation (to SLP). We thank Dr B. Leon Guerreo for his help in collecting cycad seeds. We also thank Dr Richard Smith for his financial contribution as well as his suggestions for ongoing studies of ALS-PDC. We thank Dr G. E. Kisby for supplying samples of cycasin and MAM. Professor L. T. Kurland provided much valuable advice to the authors and we dedicate this article with gratitude to his memory.


  1. Top of page
  2. Abstract
  3. Materials and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  • Ambike S. H. and Rao M. R. R. (1967) Studies on a phytosterolin from the bark of ficus Religiosa. Indian J. Pharmacol. 29, 9194.
  • Andrews J. M. (1967) Neurological disease on Guam: a review of past and present investigations. Bull. LA Neurol. Soc. 32, 3042.
  • Baraban J. M., Fiore R. S., Sanghera J. S., Paddon H. B. and Pelech S. L. (1993) Identification of p42 mitogen-activated protein kinase as a tyrosine kinase substrate activated by maximal electroconvulsive shock in hippocampus. J. Neurochem. 60, 330336.
  • Bi R., Broutman G., Foy M. R., Thompson R. F. and Baudry M. (2000) The tyrosine kinase and mitogen-activated protein kinase pathways mediate multiple effects of estrogen in hippocampus. Proc. Natl Acad. Sci. USA 97, 36023607.
  • Campbell M. E., Mickelsen O., Yang M. G., Laqueur G. L. and Keresztesy J. C. (1966) Effects of strain, age and diet on the response of rats to the ingestion of Cycas circinalis. Nutrition 88, 115124.
  • Cargnelli G., Bova S., Cannas S., Debetto P. and Luciani S. (1994) Effect of amiloride and toxic actions of ouabain in guinea-pig left atria. Eur. J. Pharmacol. 292, 103110.
  • Duncan M. W., Steele J. C., Kopin I. J. and Markey S. P. (1990) 2-amino-3-(methylamino)-propanoic acid (BMAA) in cycad flour: an unlikely cause of amyotrophic lateral sclerosis and parkinsonism-dementia of Guam. Neurology 40, 767772.
  • Kisby G. E., Ellison M. and Spencer P. S. (1992) Content of the neurotoxins cycasin (methylazoxyethanol β-d-glucoside) and BMAA (β-N-methylamino-l-alanine) in cycad flour prepared by Guam Chamorros. Neurology 42, 13361340.
  • Kurland L. T. (1988) Amyotrophic lateral sclerosis and Parkinson's disease complex on Guam linked to an environmental toxin. Trends Neurosci. 11, 5153.
  • Kurland L. T., Radhakrishnan K., Williams D. B. and Waring S. C. (1994) Amyotrophic lateral sclerosis-parkinsonism-dementia complex on Guam: epidemiological and etiological perspectives, in , Motor Neuron Disease (WilliamsA., ed.), pp. 109130 . Chapman & Hall, London.
  • Masayuki T. and Yasuo T. (1994) Structural-activity relationships of synthetic digitoxigenyl glycosides. Phytochemistry 37, 14211423.
  • Monyer H., Burnashev N., Laurie D., Sakmann B. and Seeburg P. H. (1994) Developmental and regional expression in the rat brain and functional properties of four NMDA receptor subtypes. Neuron 12, 529540.
  • Nagai A., Nakagawa E. and Choi H. B. (2001) Erythropoietin and erythropoietin receptors in human CNS neurons, astrocytes, microglia, and oligodendrocytes grown in culture. J. Neuropathol. Exp. Neurol. 60, 386392.
  • Pasqualotto B. A. and Shaw C. A. (1996) Regulation of ionotropic receptors by protein phosphorylation. Biochem. Pharmacol. 51, 14171425.
  • Peterson G. L. (1979) Review of the folin phenol protein quantitation method of Lowry, Rosebrough, Farr and Randall. Anal Biochem. 100, 201220.
  • Piccone U., Saviotti M., Pala M. and Caprari M. (1995) An unusual case of ‘catecholamine necrosis’ caused by accidental digitalis poisoning. Minerva Cardioangiol. 43, 383388.
  • Ranscht B., Clapshaw P. A. and Price J. (1982) Development of oligodendrocytes and schwann cells studied with a monoclonal antibody against galactocerebroside. Proc. Natl. Acad. Sci. USA 79, 27092713.
  • Satoh J. I. and Kim S. U. (1994) HSP72 induction by heat stress in human neurons and glial cells in culture. Brain Res. 653, 243250.
  • Satoh J. I., Tai T. and Kim S. U. (1996) Differential expression of gangliosides and galactolipids in fetal human oligodendrocytes and astrocytes in culture. Brain Res. Dev. Brain Res. 93, 172181.
  • Shaw C. A., Bains J. S., Pasqualotto B. A. and Curry K. (1999) Methionine sulfoximine shows excitotoxic actions in rat cortical slices. Can. J. Physiol. Phamacol. 77, 871877.
  • Spencer P. S., Nunn P. B., Hugon J., Ludolph A. C., Ross S. M. and Robertson R. C. (1987) Guam amyotrophic lateral sclerosis-parkinsonian-dementia linked to a plant excitant neurotoxin. Science 237, 517522.
  • Wang Y.-T. and Salter M. W. (1994) Regulation of NMDA receptors by tyrosine kinases and phsphatases. Nature 369, 233235.
  • Wilson J. M. B., Khabazian I., Wong M. C., Seyedalikhani A., Bains J.S., Pasqualotto B. A., Williams D. E., Anderson R. J., Simpson R. J. and Smith R. (2002) Behavioral and neurological correlates of ALS-parkinsonism dementia complex in adult mice fed washed cycad flour. J. Neuromol. Med. in press.