An amyloid β 42 -dependent deficit in anandamide mobilization is associated with cognitive dysfunction in Alzheimer’s disease

The endocannabinoids and their attending CB 1 cannabinoid receptors have been implicated in the control of cognition, but their possible roles in dementias are still unclear. In the present study, we used liquid chromatography/mass spectrometry to conduct an endocannabinoid-targeted lipidomic analysis of post mortem brain samples from 38 Alzheimer’s disease (AD) patients and 17 control subjects, matched for age and post mortem interval. The analysis revealed that midfrontal and temporal cortex tissue from AD patients contains, relative to control subjects, significantly lower levels of the endocannabinoid anandamide and its precursor 1-stearoyl, 2-docosahexaenoyl-sn-glycero-phosphoethanolamine-N-arachidonoyl (NArPE). No such difference was observed with the endocannabinoid 2-arachidonoyl-sn-glycerol or 15 additional lipid species. In AD patients, but not in control subjects, statistically detectable positive correlations were found between (a) anandamide content in midfrontal cortex and scores of the Kendrick’s digit copying test (P=0.004, r=0.81; n=10), which measures speed of information processing; and (b) anandamide content in temporal cortex and scores of the Boston naming test (P=0.027, r=0.52; n=18), which assesses language facility. Furthermore, anandamide and NArPE levels in midfrontal cortex of the study subjects inversely correlated with levels of the neurotoxic amyloid peptide, A β 42 , while showing no association with A β 40 levels, amyloid plaque load or tau protein phosphorylation. Finally, high endogenous levels of A β 42 in APP SWE /Neuro-2a cells directly reduced anandamide and NArPE concentrations in cells lysates. The results suggest that an A β 42 -dependent impairment in brain anandamide mobilization contributes to cognitive dysfunction in AD. language.


Introduction
The endocannabinoids are lipid-derived mediators that participate in the control of neurotransmission. The two major endocannabinoids identified so far, anandamide and 2arachidonoyl-sn-glycerol (2-AG), are released on demand through activity-or receptordependent hydrolysis of membrane phospholipid precursors, and act on presynaptic CB 1type cannabinoid receptors to inhibit neurotransmitter release (Piomelli et al., 2007;Katona and Freund, 2008). They regulate synaptic plasticity in various regions of the brain through both short-term and long-term mechanisms, which include depolarization-induced suppression of inhibition at hippocampal GABAergic synapses and long-term depression at cortical or accumbal glutamatergic synapses (Hashimotodani, 2007;Chevaleyre et al., 2006;Mato et al., 2008).
Anandamide and 2-AG are produced through two separate enzymatic pathways Piomelli, 2003). Anandamide formation starts with the transfer of arachidonic acid from phosphatidylcholine to phosphatidylethanolamine (PE), which generates a diverse group of N-arachidonoyl-substituted PE species (NArPE) (Piomelli, 2003). Anandamide is released by the hydrolysis of NArPE, which requires either a NAPEspecific phospholipase D (PLD) (Okamoto et al., 2004) or the sequential actions of phospholipase C and PTPN22 phosphatase (Liu et al., 2006). After release, anandamide is internalized by neurons and glia and then degraded by intracellular fatty acid amide hydrolase (FAAH) (Piomelli, 2003).
Evidence indicates that endocannabinoid signaling is involved in the regulation of normal cognition. Pharmacological or genetic blockade of CB 1 receptors in mice impairs short-and long-term memory extinction (Lutz, 2007;de Oliveira Alvares et al., 2008;Marsicano et al, 2002) while disruption of FAAH-mediated anandamide hydrolysis enhances extinction in mice (Varvel et al., 2007) and improves memory consolidation in rats (Mazzola et al., 2009). Additionally, mutant mice lacking CB 1 receptors display an accelerated cognitive decline with aging (Bilkei-Gorzo et al., 2005), which is accompanied by alterations in spine morphology (Ballesteros-Yáñez et al., 2007). Deficits in endocannabinoid signaling might also contribute to dementias. CB 1 and CB 2 cannabinoid receptors were found to be associated with amyloid β-protein (Aβ) plaques in post mortem brain tissue from subjects with Alzheimer's disease (AD), (Ramirez et al., 2005). Moreover, elevated expression levels of CB 2 receptors and FAAH were observed in immune cells surrounding senile plaques in subjects with AD or Down's syndrome (Núñez et al., 2008;Centonze et al., 2007;Benito et al., 2007). Lastly, pharmacological agents that enhance endocannabinoid activity were shown to exert beneficial effects in animal models of Aβ-induced toxicity. Cannabinoid agonists prevented Aβ-triggered microglial activation and neurotoxicity in primary cell cultures, suggesting that cannabinoid receptor activation may reduce neuroinflammation (Ehrhart et al., 2005;Ramirez et al., 2005). Consistent with these results, the anandamide reuptake inhibitor VDM-11 reversed hippocampal damage and loss of memory retention in rodents treated with Aβ 42 peptide (van der Stelt et al., 2006). Although CB 1 receptor expression was not found to be altered in the post mortem cortex of AD patients (Lee et al., 2010), the majority of available data collectively suggests an involvement of the endocannabinoid system in the cognitive deterioration that accompanies AD. To test this hypothesis, here we utilized liquid chromatography/mass spectrometry (LC/MS) to characterize the endocannabinoid-related lipidome in the cortex and cerebellum of subjects with AD and non-demented control subjects, closely matched for age and post mortem interval.

Research subjects
We analyzed frozen samples of brain tissue from a total of 17 non-demented control subjects and 38 pathologically confirmed subjects with AD (males/females: control subjects, 10/7; subjects with AD, 20/18), provided by the Institute for Memory Impairments and Neurological Disorders and the Alzheimer's Disease Research Center at the University of California, Irvine. Three brain areas were selected for analysis: midfrontal cortex (Broadman area 9), temporal cortex (Broadman area 20) and cerebellum.  (Folstein et al., 1975), were accessible for 8 control subjects (mean score ±SEM = 28.4±0.6; assessed 45.9±9.3 months before death) and 20 subjects with AD (mean score±SEM = 11.4±1.6; assessed 10.5±2.2 months before death). Boston Naming test scores (Kaplan et al., 1983) and Kendrick Digit Copy test scores (Kendrick, 1985) were also available (Supplementary Table 1). Medications taken were monitored in part of the patients (n=11) (Supplementary Table 2).

Cell cultures
We cultured Neuro-2a and APP SWE cells (kindly provided by Drs. Seong-Hun Kim and Sangram S. Sisodia, University of Chicago, IL) at 37°C with 5% CO 2 in Dulbecco's modified Eagle's medium (Invitrogen, Carlsbad, CA) supplemented with 10% fetal bovine serum and antibiotics (100 unit/ml Penicillin G and 100 μg/ml Streptomycin; Invitrogen). For the maintenance of APP SWE cells, G418 (0.4 mg/ml, Invitrogen) was included in the media. After 72 hours of plating into 100-mm culture dishes, cells were collected for analyses.

Lipid extractions from cells in cultures
Cells were washed with ice-cold phosphate-buffered saline (PBS) and scraped into 1 ml of methanol/water (1:1, vol:vol) containing the internal standards listed above. Protein concentrations were measured using the BCA protein assay (Pierce, Rockford, IL). Lipids were extracted with chloroform/methanol (2:1, vol:vol; 1.5 ml). The organic phases were collected, dried under N 2 and dissolved in chloroform/methanol (1:3, vol:vol) for LC/MS analyses.

LC/MS analyses
Monoacylglycerols (MG) and fatty acid ethanolamides (FAE) -We used an Agilent 1100-LC system coupled to a 1946D-MS detector equipped with an electrospray ionization (ESI) interface (Agilent Technologies, Inc., Palo Alto, CA). Lipids were separated on a reversedphase XDB Eclipse C18 column (50×4.6 mm i.d., 1.8 μm, Zorbax, Agilent Technologies). They were eluted with a gradient of methanol in water (from 85% to 90% methanol in 2.0 min and 90% to 100% in 3.0 min) at a flow rate of 1.5 ml/min. Column temperature was kept at 40°C. MS detection was in the positive ionization mode, capillary voltage was set at 3 kV and fragmentor voltage was 120 V. N 2 was used as drying gas at a flow rate of 13 l/ min and a temperature of 350°C. Nebulizer pressure was set at 60 PSI. For quantification purposes, we monitored the sodium adducts of the molecular ions [M+Na] + in the selected ion-monitoring (SIM) mode, using [ 2 H 8 ]-2-AG (mass-to-charge ratio, m/z 409) and [ 2 H 4 ]anandamide (m/z 370) as internal standards. Non-esterified fatty acids -We used a reversed-phase XDB Eclipse C18 column (50×4.6 mm i.d., 1.8 μm, Zorbax, Agilent Technologies) eluted with a linear gradient from 90% to 100% of A in B for 2.5 min at a flow rate of 1.5 ml/min with column temperature at 40°C. Mobile phase A consisted of methanol containing 0.25% acetic acid and 5 mM ammonium acetate; mobile phase B consisted of water containing 0.25% acetic acid and 5 mM ammonium acetate. ESI was in the negative mode, capillary voltage was set at 4 kV and fragmentor voltage was 100 V. N 2 was used as drying gas at a flow rate of 13 l/min and a temperature of 350°C. Nebulizer pressure was set at 60 PSI. We monitored deprotonated molecular ions [M-H] − in the SIM mode and [ 2 H 8 ]-arachidonic acid (m/z 311) was used as an internal standard. N-acylphosphatidylethanolamine (NAPE) -NAPE species, including NArPE, were analyzed as previously described ). Briefly, they were separated by LC using an 1100 system (Agilent Technologies) equipped with a Poroshell 300 SB C18 column (2.1 × 75 mm inner diameter, 5 μm; Agilent Technologies) maintained at 50°C. A linear gradient of methanol in water containing 5 mM ammonium acetate and 0.25% acetic acid (from 85% to 100% of methanol in 4 min) was applied at a flow rate of 1 ml/min. MS analyses were conducted using an Ion Trap XCT (Agilent Technologies) set in the negative mode. Capillary voltage was 4.5 kV, with skim1 at −40 V and capillary exit at −151 V. N 2 was used as drying gas at a flow rate of 12 l/min, with temperature at 350°C and nebulizer pressure at 80 psi. Helium was used as the collision gas. Extracted ion chromatograms were used to quantify each NAPE precursor ion by monitoring the characteristic lyso-NAPE product ions in MS 2 using 1,2-dipalmitoyl-sn-glycero-3-phosphoethanolamine-Nheptadecanoyl (m/z 942.8 > 704.8) as an internal standard.

In vitro FAAH activity assay
Cells were washed and homogenized in ice-cold 50 mM Tris-HCl, pH 7.5. Homogenates were centrifuged at 1000 × g for 10 min at 4 °C and supernatants were collected for in vitro FAAH assays. Protein concentration was determined using a BCA protein assay kit (Pierce). We measured FAAH activity at 37°C for 30 min in 0.5 ml of Tris buffer (50 mM, pH 7.5) containing fatty acid-free bovine serum albumin (0.05%, w/v), the cell homogenates (50 μg), 10 μM anandamide, and anandamide[ethanolamine-3 H] (10,000 cpm, specific activity 60 Ci/mmol; American Radiolabeled Chemicals, St. Louis, MO). The reactions were stopped with chloroform/methanol (1:1, 1 ml) and radioactivity was measured in the aqueous layers by liquid scintillation counting.

Statistical analyses
Results are expressed as means ± SEM. The significance of differences among groups was evaluated using the two-tailed Student's t-test and differences were considered significant if P<0.05. Pairwise correlation between variables was assessed by Pearson and confirmed by Spearman rank. Analyses were conducted using GraphPad Prism (GraphPad Software, San Diego, CA).

Anandamide mobilization is impaired in AD
Demographic and clinical information on the subjects included in the present study is provided in Supplementary Tables 1 and 2. There was no effect of age, gender, post mortem interval or medication on the levels of anandamide (Supplementary Figs. 1 and 2) or other endocannabinoid-related lipids targeted by our analysis (Table I and data not shown). By contrast, there was a statistically detectable difference between control subjects and AD patients in the levels of anandamide in samples of midfrontal cortex (P=0.047, Fig. 1A) and temporal cortex (P=0.034, Fig. 1B). No such difference was found in cerebellum (Fig. 1C) or with other lipid species, including 2-AG (Fig. 1D), palmitoylethanolamide (Fig. 1E) and arachidonic acid (Fig. 1F) (Table 1 and Supplementary Tables 3 and 4). The results indicate that anandamide content is selectively lowered in midfrontal and temporal cortex of AD patients, relative to control subjects.
To determine whether this decrease might be due to a defect in anandamide production, we quantified by LC/MS the predominant NArPE species found in the human brain: 1-stearoyl, 2-docosahexaenoyl-sn-glycero-phosphoethanolamine-N-arachidonoyl (Astarita et al., 2008). We found a statistically detectable difference between control subjects and AD patients in NArPE content of midfrontal cortex (P=0.0019, Fig. 1G) and temporal cortex (P=0.0004, Fig. 1H), but not cerebellum (P=0.47, Fig. 1I). Statistical inspection of the data revealed a significant correlation between the levels of NArPE and those of anandamide (P=0.0009 and P<0.0001, r=0.44 and 0.55 in the midfrontal and temporal cortex, respectively, Supplementary Fig. 3), which confirmed the expected precursor-product relationship between the two molecules (Astarita et al., 2008).
The findings reported above suggest that AD is associated with a deficit in brain anandamide production. In addition, we also observed a trend toward elevation of FAAH mRNA levels in temporal cortex of AD patients, but such trend did not reach statistical significance ( Supplementary Fig. 4). This observation is consistent with the results of previous studies, which have reported increased FAAH expression in neuritic plaqueassociated astrocytes and microglia of AD and Down's syndrome patients (Benito et al, 2003;Núñez et al, 2008).

Anandamide levels correlate with cognitive performance
The functional significance of the deficit in anandamide mobilization observed in the cortex of AD patients was explored using correlation analyses. A highly significant positive correlation was found between anandamide content in midfrontal cortex of AD patients and their performance in the Kendrick's digit copy test (P=0.004 and 0.046, Figs. 2A and 2B), which measures psychomotor speed. By contrast, no correlations were found between anandamide levels in the midfrontal cortex of the patients and scores of either the Boston naming test (Fig. 2C), which assesses the ability to name pictures of objects, or the MMSE test ( Supplementary Fig. 5A), a measure of global cognitive function. On the other hand, anandamide levels in temporal cortex of AD patients were positively associated with scores of the Boston naming test (P=0.027, Fig. 2F), but not the Kendrick's digit copy test (Figs. 2D and 2E). Levels of 2-AG or palmitoylethanolamide displayed no correlation with any of those tests (Supplementary Fig. 6). These correlations do not prove causation, but do suggest that impaired anandamide mobilization in midfrontal and temporal cortex accompanies a deterioration of specific cognitive abilities in subjects with AD.

Aβ 42 disrupts anandamide mobilization
To determine whether neurotoxic Aβ peptides can impair anandamide mobilization, we first asked whether cortical levels of Aβ peptides are statistically correlated with those of anandamide in our subject group. As expected, midfrontal cortex tissue from AD patients contained higher concentrations of both SDS-extractable (soluble) and SDS-insoluble forms of Aβ 42 and Aβ 40 , compared to tissue from non-demented control subjects (Figs. 3A and 3D). Importantly, we found a statistically detectable negative correlation between Aβ 42 and anandamide content (P=0.0016 and 0.014, Figs. 3B and 3E). Levels of the anandamide precursor, NArPE, were also significantly correlated with insoluble Aβ 42 content (P=0.0034, Fig. 3C), while a trend toward correlation, which did not reach statistical significance, was observed with soluble Aβ 42 (Fig. 3F). No significant correlations were found between levels of Aβ 40 and anandamide or NArPE (Supplementary Fig. 7). A separate analysis for control and AD brains for the correlations also indicated that levels of SDS-insoluble Aβ 42 significantly correlate with anandamide and NArPE in AD patients (P=0.03 and 0.04, respectively) (Supplementary Tables 5). In addition, anandamide levels in midfrontal cortex did not correlate with amyloid plaque load, neurofibrillary tangles or ApoE4 genotype ( Supplementary Fig. 8), which is suggestive of a specific link between Aβ 42 and anandamide mobilization.
To further explore this link, we stably overexpressed the familial early-onset AD (FAD)associated Swedish mutant form of amyloid precursor protein (APP SWE ) in mouse Neuro-2a cells (Thinakaran et al., 1996). Overexpression of APP SWE and accumulation of cellassociated Aβ 40 and Aβ 42 were confirmed using western blot and ELISA assays, respectively (Figs. 4A and 4B). These changes were associated with a significant reduction in the levels of anandamide (Fig. 4C) and those of 1-O-octadecenoyl, 2-octadecenoyl-snglycero-phosphoethanolamine-N-arachidonoyl (NArPE # ), which we have identified by LC/ MS as the most abundant NArPE species in Neuro-2a cells (Fig. 4D). Additionally, whereas NAPE-PLD mRNA levels were not affected (Fig. 4E), FAAH mRNA levels were significantly increased in APP SWE cells compared to control cells (Fig. 4F). Such increase was accompanied by elevated expression of FAAH protein (Fig. 4G) as well as in vitro FAAH activity (Fig. 4H). APP SWE overexpression had no effect on cell viability (data not shown). These findings provide direct evidence that abnormally high levels of endogenous Aβ 42 alter anandamide mobilization in neuronal cells.

Discussion
The objective of the present study was to investigate the link, suggested by previous reports, between brain endocannabinoid signaling and AD (Ramirez et al., 2005;Núñez et al., 2008;Centonze et al., 2007;Benito et al., 2007). Our targeted lipidomic analyses of human brain tissue provide new evidence in support of such a link by showing that levels of the endocannabinoid anandamide are significantly lower in midfrontal and temporal cortex of subjects with AD than of non-demented subjects closely matched for age and post mortem interval. To identify potential mechanisms contributing for this AD-associated alteration, we quantified the anandamide precursor, NArPE, in the same brain samples (Cadas et al., 1996;Cadas et al., 1997;Astarita et al., 2008). The results indicate that NArPE content is also markedly reduced in the cortex of AD patients, which is suggestive of a deficit in anandamide production. While limited by the use of post mortem tissue, in which endocannabinoid levels are likely to be altered (Schmid et al., 1995), the functional significance of our results is underscored by the identification of highly significant positive correlations between cortical anandamide content and the patient's performance in cognitive measures of psychomotor speed and language.
Synaptic deterioration is a hallmark of AD pathology (Cotman and Anderson, 2000). It does not appear, however, that the changes in cortical anandamide levels documented here were the consequence of a generalized loss of synaptic lipids. Arguing against this possibility, we found that among the 16 endocannabinoid-related lipid species targeted by our analyses, only anandamide and its precursor NArPE were altered in AD. In particular, we observed no significant changes in the levels of another endocannabinoid, 2-AG, a polyunsaturated fatty acid, arachidonic acid, and a lipid amide, palmitoylethanolamide. These lipid molecules provide a sensitive indicator of synaptic lipid degradation because they are released in substantial amounts during ischemic brain damage (Panikashvili et al., 2001;Schabitz et al., 2002). The results of prior lipid-profile studies of human brain strengthen the idea that AD is associated with a restricted set of lipid abnormalities rather than a generalized loss of neuronal lipids (Soderberg et al., 1991;Skinner et al., 1993;Prasad et al., 1998;Guan et al., 1999;Lukiw et al., 2005;Fraser et al., 2010;Astarita et al., 2010). Also supportive of this conclusion are two additional findings. First, van der Stelt et al have shown that injections of neurotoxic Aβ 42 into the rat hippocampus cause a transient increase in 2-AG levels followed by a more persistent decrease in anandamide levels (van der Stelt et al., 2006). Second, we show here that cortical levels of anandamide correlate with the patients' performance in cognitive tests of psychomotor speed and linguistic ability, but not with amyloid plaque burden or tau-protein hyperphosphorylation, two features of AD neuropathology that are linked to neuronal cell death (Haass and Selkoe, 2007). Notably, animal experiments show that altered endocannabinoid signaling in hippocampus might also contribute to memory loss following Aβ 42 administration (van der Stelt et al., 2006;Mazzola et al., 2003).
Neurotoxic amyloid Aβ peptides are thought to play a key role in the pathogenesis of AD (Gandy, 2005). It is significant, therefore, that cortical levels of anandamide and its precursor NArPE were inversely correlated with those of the highly amyloidogenic peptide, Aβ 42 (Iwatsubo et al., 1994;McGowan et al., 2005), but not with those of the less toxic species, Aβ 40 (Dahlgren et al., 2002). Though further research is needed to determine if this correlation represents cause and effect, our experiments with APP SWE -overexpressing Neuro-2a cells provided direct evidence that pathological accumulation of Aβ 42 can disrupt anandamide mobilization, while having no overt effect on cell viability. These studies suggest that excessive Aβ 42 can simultaneously impact both anandamide production, by curtailing availability of the anandamide precursor NArPE, and anandamide degradation, by increasing expression of the anandamide-hydrolyzing enzyme FAAH. Similar alterations have been observed in brain tissue of subjects with AD and Down's syndrome (Benito et al, 2003;Núñez et al, 2008;present study), supporting the possibility that an Aβ 42 -linked mechanism may impair anandamide mobilization in AD and remove a protective influence of this endocannabinoid messenger against neural toxicity and inflammation (Ehrhart et al., 2005;Ramirez et al., 2005;van der Stelt et al., 2006). The molecular events engaged by Aβ 42 to affect anandamide mobilization were not investigated in the present study. Interesting parallels are offered, however, by prior findings showing that pro-inflammatory stimuli down-regulate lipid amide biosynthesis in innate immune cells (Solorzano et al, 2009) and nerve injury enhances FAAH expression in peripheral sensory neurons (Lever et al., 2009). While additional studies are needed to fill this knowledge gap, it is tempting to speculate that pharmacological agents that strengthen anandamide signaling, such as FAAH inhibitors, might be beneficial to improve cognition in AD.