Introduction
Decades of Alzheimer’s disease (AD) research have been grounded on the so called “amyloid cascade hypothesis”, which originally placed amyloid precursor protein (APP) mismetabolism and subsequent Aβ aggregation (i.e., fibrillation) as the initial trigger responsible for instigating further pathological events (i.e., tauopathy, synaptic damage, and neuronal death) [
49,
52,
97]. However, amyloid deposits were later shown to correlate poorly with cognitive decline and to be disconnected from Aβ-induced toxicity [
29,
68,
72,
85]. On the other hand, characterization of soluble Aβ structures led to the discovery of Aβ derived diffusible ligands (ADDLs) or oligomeric Aβ [
63]: extremely neurotoxic species that strongly correlate with synaptic impairment and parallel cognitive decline in animal models and humans [
11,
36,
53,
62,
63,
68,
120,
122,
123]. Importantly, it has been shown that oligomeric Aβ species are both necessary and sufficient to disrupt cognitive function in vivo [
21,
64,
107]. These findings led to a “revised” amyloid cascade hypothesis where diffusible oligomeric Aβ replaced fibrillar Aβ as the central neurotoxic event driving AD pathogenesis [
48,
98].
The E693Δ (Osaka) mutation in APP, which was found in Japanese pedigrees, causes familial AD by enhancing Aβ oligomerization in the absence of deposits of amyloid plaques [
116]. The mutant Aβ peptide, which lacks glutamate-22 (E22Δ), forms abundant oligomers in vitro and causes endoplasmic reticulum stress-induced apoptosis in cultured cells [
83]. When injected into rat cerebral ventricle, synthetic mutant Aβ E22Δ peptide inhibits hippocampal long-term potentiation more potently than wild-type (WT) peptide [
116]. Exogenously applied Aβ E22Δ peptide induces dose-dependent loss of synapses in mouse hippocampal slices [
112]. In addition, APP
E693Δ transgenic mice (APP
OSK) showed intraneuronal accumulation of Aβ oligomers, synapse loss, memory impairment, and significant neuronal loss at 24 months of age [
115]. Thus, APP
OSK mice successfully recapitulate Aβ neurotoxicity in the absence of amyloid plaques.
When Alois Alzheimer first described the disease over 100 years ago, he identified abnormal protein deposits as well as adipose saccules (lipid inclusions) in the brains of his patients [
4]. These observations suggested a possible relation between AD and lipid imbalance, which was established decades later when the strongest genetic risk factor in AD was linked to apolipoprotein E (apoE), the major lipid transporter in the CNS. Dysregulation of multiple lipid families has been linked to AD, including alterations in the levels of sulfatide, plasmalogen ethanolamine glycerophospholipid, cholesterol, ceramide, and fatty acids (FAs) [
15,
18,
19,
23,
26,
41,
43,
44,
46].
FAs and their metabolites are of particular relevance, given that they participate in processes involved in the pathogenesis of AD, including synaptic plasticity, inflammation, cerebrovascular function, and oxidative stress [
74,
88,
89,
106]. FAs are released from phospholipids by phospholipase A
2 (PLA
2) [
61], a family of enzymes that catalyze the cleavage of FAs from the
sn-2 position of phospholipids. These enzymes are not only important for maintenance of cellular membrane phospholipids, they also play a key role in regulating the release of signaling molecules like arachidonic acid (AA) and docosahexaenoic acid (DHA), important precursors for lipid-derived modulators of cell signaling and inflammatory processes. Given that phospholipids within CNS membranes are enriched in polyunsaturated fatty acids (PUFAs) [
110] and that the
sn-2 position is mostly constituted with unsaturated FAs, PLA
2 cleavage activity within the brain results in accumulation of lysophospholipids and unsaturated FAs [
93,
96,
134].
In the mammalian system, more than 19 different isoforms of PLA
2 have been identified, and different PLA
2s have been shown to participate in physiological events related to cell injury, inflammation, and apoptosis [
24,
81]. Research to understand PLA
2s in the CNS has focused on 3 PLA
2 isoforms: the group IV Ca
2+-dependent cytosolic PLA
2 (cPLA
2), which has been strongly associated with AD (reviewed in [
94]) [
22,
95,
108,
109,
111,
113]; the group VI Ca
2+-independent PLA
2 (iPLA
2), which has been proposed to account for >70% of brain PLA
2 activity [
132] and is highly enriched in AD-affected brain regions (i.e., cortex and hippocampus) [
87]; and the group II secretory PLA
2 (sPLA
2), which has also been linked to AD more recently [
14,
80].
Importantly, the activity of cPLA
2 has been shown to be tightly regulated by multiple mechanisms. First, cPLA
2 becomes activated after translocating to the plasma membrane from the cytosol [
35]. Although Ca
2+ is not necessary for cPLA
2 catalytic activity, nanomolar Ca
2+ concentrations are needed for its binding to the membrane [
32]. Second, it is well-established that phosphorylation of cPLA
2α at multiple sites (Ser505 and Ser515) stimulates its catalytic activity [
30,
34,
50,
60,
66]. In vitro work has revealed that protein kinase C (PKC) plays an important role in mediating cPLA
2 phosphorylation and AA release in murine astrocytes through both MAPK-dependent and MAPK-independent pathways [
131]. Third, cPLA
2 regulation via protease-mediated cleavage has also been documented [
1,
5,
6,
30,
40,
119,
128], although this regulation seems to occur only under apoptotic and/or necrotic conditions. Conflicting data have been reported regarding the effects of cPLA
2 proteolysis which has been found to both activate [
30,
40,
128] and inhibit [
1,
5,
6,
119] its activity. Notably, cPLA is highly specific to AA cleavage/release [
28,
100,
101] and to PC [
81]. In fact, cPLA
2α-deficient mice fail to generate AA metabolites after brain injury [
9,
61], thus cPLA
2α seems to be the most relevant cPLA
2 in the brain.
Like cPLA
2, the activity of iPLA
2 also seems to be tightly controlled. First, it has been well-established that iPLA
2 is inhibited by calmodulin and activated by Ca
2+ release from ER where calcium influx factor (CIF) has been proposed to displace inhibitory calmodulin [
104,
129,
130]. Second, it has been proposed that PKC mediates phosphorylation of iPLA
2 (directly and/or indirectly) promoting its activity [
75]. Third, caspase-3-dependent cleavage and activation of iPLA
2 have been documented [
136]. Notably, murine studies have reported expression of an 80-kDa iPLA
2 isoform (iPLA
2β encoded by the
PLA2G6 gene) in brain tissue [
132]. iPLA
2β has been shown to be physiologically and clinically relevant, as demonstrated by characterization of iPLA
2β-KO mice which model neurodegeneration with brain iron accumulation [
70,
102] and by the fact that mutations in the
PLA2G6 gene lead to two childhood neurologic disorders [
39,
56,
78].
Although multiple lipid classes and lipid cleavage enzymes have been associated to AD, whether lipid dysregulation plays a causative or epiphenomal role in the disease remains largely unknown. In the current study, we took advantage of the APPOSK mouse model where AD-like pathology and neurodegeneration occur in the absence of amyloid plaques, and demonstrated that oligomeric amyloid-beta (Aβ) induces accumulation of free PUFAs and lysophosphatidylcholine by activation of brain cPLA2 and iPLA2 within myelin-rich and pyramidal neuron-rich regions, respectively, via MAPK-mediated phosphorylation in a PKC-independent manner.
Materials and methods
Mice
Brain tissue from 12 and 24 month old APP
OSK-Tg, APP
WT-Tg, and non-Tg mice (
n = 4/genotype including an equal mix of male and female mice) was kindly obtained from Dr. Takami Tomiyama, Associate Professor from the Osaka City University. As previously described, three lines of APP-Tg lines have been established for APP
WT and APP
OSK mice with high (L1), low (L2), and intermediate (L3) human expression of the transgene [
115]. All the studies were performed using L1 APP-Tg lines. It is important to note that APP
WT L1 mice express higher levels of human APP (2-fold) than APP
OSK L1 mice do [
115].
Mouse cerebrum tissue was sub-dissected by removal of olfactory lobe, cerebellum, brain stem, and colliculus from each hemibrain. Frozen cerebrum samples were weighed, lyophilized, pulverized, and homogenized in 500 μl of ice-cold diluted phosphate-buffered saline (0.1X PBS) on a cooling tissue homogenizer (Cryolys Precellys Evolution Homogenizer). Protein assays on individual homogenates were performed using a BCA protein assay kit (Pierce, Rockford, IL, USA). Lipids were extracted by a modified procedure of Bligh and Dyer extraction as described previously [
16,
17] in the presence of internal standards which were added based on total protein content of the sample.
Mass spectrometric analysis of lipids
A triple-quadrupole mass spectrometer (Thermo Scientific TSQ Vantage, CA, USA) equipped with a Nanomate device (Advion Bioscience Ltd., NY, USA) and Xcalibur system software was used as previously described [
47,
133]. Diluted lipid extracts were directly infused into the ESI source through a Nanomate device [
47]. Typically, signals were averaged over a 1-min period in the profile mode for each full scan MS spectrum. For tandem MS, a collision gas pressure was set at 1.0 mTorr, but the collision energy varied with the classes of lipids as described previously [
45,
133]. Similarly, a 2- to 5-min period of signal averaging in the profile mode was employed for each tandem MS mass spectrum. All full and tandem MS mass spectra were automatically acquired using a customized sequence subroutine operated under Xcalibur software. Data processing including ion peak selection, baseline correction, data transfer, peak intensity comparison,
13C deisotoping, and quantitation were conducted using a custom programmed Microsoft Excel macro as previously described [
133] after considering the principles of lipidomics [
125].
Elisa
The levels of Aβ oligomers were quantified by direct ELISA with anti-human amyloid-β E22P (11A1) mouse IgG monoclonal antibody (IBL, Japan) at 1 μg/ml as previously described [
13,
118]. Briefly, PBS supernatants (at a concentration of 5000 μg/ml of total protein) were diluted 6-fold in sodium bicarbonate pH 9.6 (0.5X ELISA Plate Coating buffer, Alpha Diagnostic International, TX) and allowed to coat ELISA plates at 50 μl/well. After incubation with HRP-conjugated anti-mouse IgG, 11A1 immunoreactivity was detected using 3,3′,5,5′-Tetramethylbenzidine (TMB-1, Alpha Diagnostic International, TX). Reactions were stopped with diluted sulfuric acid (1X Stop Solution, Alpha Diagnostic International, TX).
Western blot analysis
Pulverized cerebrum tissues were homogenized in 1X NP40 on a cooling tissue homogenizer (Cryolys Precellys Evolution Homogenizer). NP40 homogenates were centrifuged at 12,300 rpm for 20 min at 4 °C and supernatants were run into NuPage 4–12% Bis-Tris (Life Technologies, NY) under reducing conditions. Samples were normalized based on total protein content, which was estimated by the BCA protein assay. Western blot analyses were performed using antibodies against cPLA2 (sc-454 and sc-376,636, Santa Cruz Biotechnology -SCB-), phospho-cPLA2 -S505- (2831, Cell Signaling Technology -CST-), iPLA2 (sc-376,563, SCB; and NBP1–81586, Novus), MAPK p42/p44 (4695, CST), phospho-MAPK p42/p44 -T202/Y204- (4370, CST), MAPK p38 (8690, CST), phospho-MAPK p38 (rabbit polyclonal, CST), SAPK/JNK (9252, CST), active JNK (V7931, Promega), PKCα (2056, CST), PKCδ (9616, CST), PKCλ (610,207, BD Biosciences), phospho-PKC pan -βII Ser660- (9371, CST), phospho-PKCα/βII -T638/641- (9375, CST), phosphor-PKCδ -T505- (9374, CST), phospho-PKCζ/λ -T410/403- (9378, CST), CaMK2 (sc9035, SCB), pCaMK2 (sc-12,886-R, SCB), β-Tubulin (2146, CST), GAPDH (MAB374, Millipore), VDAC (4866, CST). Relative intensities were quantified using ImageJ software.
MALDI imaging
Matrix-assisted laser desorption/ionization (MALDI) imaging of fatty acids was carried out as previously described [
124]. Briefly, fresh frozen brain from adult C57BL/6 J WT mouse was cryosectioned at 10-μm thickness. Brain slices were transferred onto the conductive side of indium tin oxide (ITO) slides and desiccated in vacuum for 30–60 min. After drying, N-(1-naphthyl) ethylenediamine dihydrochloride matrix was applied by the Bruker ImagePrep device (Bruker Daltonics, Bremen, Germany). MALDI mass spectra were acquired in the negative ion mode using a reflectron geometry MALDI-TOF mass spectrometer (Ultraflextreme; Bruker Daltonics) equipped with a neodymium-doped yttrium aluminum garnet (Nd:YAG)/355-nm laser as the excitation source. Imaging data were analyzed using FlexImaging v3.0 and BioMap v3.8. Ion images were generated with a bin width of ±0.2 Da. The normalization method was total ion count (TIC).
Immunofluorescence
Mouse brains were dissected, fixed in 4% paraformaldehyde, cryoprotected, embedded in OCT, and frozen. Cryostat brain sections (8 μm) were mounted on positively charged slides. Tissue-containing slides were incubated with phospho-cPLA2 -S505- (2831, CST), iPLA2 (NBP1–81586, Novus), and NeuN (NAB377, Millipore) primary antibodies overnight at 4 °C and incubated with secondary antibody (Goat anti-mouse Alexa Fluor® 555, Goat anti-rabbit Alexa Fluor® 647) for 1 h at room temperature, followed by the addition of DAPI-containing mounting media (Vectashield, Vector Laboratories). Images were taken using 20× and 40× objectives on a Nikon A1R VAAS inverted confocal microscope and analyzed using NIS-Elements imaging software (Nikon).
Statistical analysis
Quantitative data were normalized to protein content and were presented as the means ± SE. Differences between mean values were determined by unpaired Student’s t test (one time point analysis comparing the abundance of specific lipid species or protein between the 3 different genotypes) using GraphPad Prism software.
Discussion
Seeking to better understand the role of fatty acid metabolism in AD and to unravel the mechanisms underlying its disruption we took advantage of the powerful technology of multidimensional mass spectrometry-based shotgun lipidomics (MDMS-SL) pioneered by our laboratory. At the same time, attempting to dissect if fatty acid dysregulation is linked to fibrillar and/or soluble Aβ accumulation, we took advantage of the APPOSK mouse model where AD-like pathology and neurodegeneration occur as a consequence of high levels of soluble oligomeric Aβ in the absence of amyloid plaques. Importantly, besides non-Tg control mice, our studies also included APPWT transgenic controls; this experimental design allowed us to distinguish between the effects of APP overexpression and those of oligomeric Aβ accumulation. To the best of our knowledge, this is the first study to demonstrate that (1) dysregulation of fatty acid metabolism in the context of AD occurs independently of amyloid plaques, (2) APP overexpression on its own induces an accumulation of unsaturated NEFAs (particularly AA, DHA, and OA) and lysoPCs while soluble oligomeric Aβ further exacerbates this accumulation of unsaturated FAs/lysoPCs, (3) cPLA2/AA and iPLA2/DHA accumulate in different and opposite brain regions, (4) AD-related fatty acid dysregulation is induced by PLA2 activation via MAPK-mediated phosphorylation in a PKC-independent manner.
Thanks to the fact that disruption of fatty acid metabolism results in specific signatures within the lipidome (depending on whether the alterations are induced through biosynthesis and/or degradation pathways), our lipidomics approach enabled us to not only demonstrate that both APP overexpression and high oligomeric Aβ content lead to significant and additive increases in unsaturated NEFAs within the brain of aged mice, but also to gain insights into the mechanisms underlying this disruption of FA metabolism. The lipidomics signature obtained strongly indicated that the accumulation of NEFA occurred as a consequence of increased FA cleavage. Specifically, analysis of the two major lysophospholipids classes revealed a dramatic accumulation of lysoPCs. Detailed characterization of specific lysoPC species indicated that FA cleavage in the context of AD seems to be induced by increased PLA2 activity as well as oxidative stress (revealed by an accumulation of 4-HNE and sn-2 lysoPCs). We validated this lipidomics-based hypothesis by analyzing the most abundant PLA2 enzymes in the brain by WB and provided evidence supporting a model in which increased PLA2 activity is mediated by phosphorylation of cPLA2α and iPLA2β.
These results are in agreement with multiple previous reports linking high AA content and cPLA
2 activity/phosphorylation to AD (reviewed in [
94]) [
22,
95,
108,
109,
111,
113]. Furthermore, two separate groups have recently reported increased free DHA levels in human AD brains [
82,
105]. In addition, the proportion of phospholipid-bound DHA has been reported to be decreased in AD brains [
25], which is also consistent with increased FA cleavage rates. The fact that iPLA
2 has been shown to preferentially cleave/release DHA from brain phospholipids [
37,
69] together with our data from AD mice and human data from other labs showing increased levels of free DHA, strongly implicate iPLA
2 as a potential driver of this accumulation. To the best of our knowledge we are the first ones to propose an association between iPLA
2β and AD.
Taken together our results strongly suggest that both PLA2-mediated accumulation of free PUFAs and oxidative stress drive AD-related disruption of lipid metabolism. Furthermore, we think that oligomeric Aβ-induced NEFA accumulation might be associated with the “adipose inclusions” described by Alois Alzheimer more than a century ago.
Opposite brain spatial distribution between cPLA2/AA and iPLA2/DHA
In vitro studies have shown that cerebral microvascular endothelium and astrocytes can produce DHA and AA [
76,
77]; in contrast, neurons cannot produce PUFAs but get enriched with PUFAs if they are co-cultured with astrocytes and endothelial cells. Interestingly, MALDI-MS imaging analysis revealed that AA and DHA accumulate within different and opposite brain regions. We found that although free DHA is detected throughout the brain, it accumulates most strongly within cortical and hippocampal regions, both of which are rich in pyramidal neurons/dendritic spines and are severely affected in AD. Consistent with this observation, previous studies have reported that 50% of the weight of neuronal plasma membrane is composed by DHA [
103,
121]. Our results suggest that even under physiological conditions, there is a high exchange between free and lipid-bound DHA, presumably due to the high levels of plasma membrane remodeling that occur within dendritic spines (which are particularly enriched in pyramidal neurons). On the other hand, AA levels were strongest along bundles of nerve fibers and moderate within thalamic and hypothalamic regions, while cortical and hippocampal regions showed negligible levels of AA. It is reasonable to speculate that the opposite localization of free AA and DHA within the brain could be evolutionarily related to their opposite roles as mediators of pro- and anti-inflammatory signaling pathways.
Given that epidemiological research has linked high DHA consumption with a lower risk of AD [
79] and animal studies have reported a reduction of amyloid, tau, and neuritic pathology with oral intake of DHA [
12,
38,
65], it could seem paradoxical that AD brains accumulate free DHA. However, it is important to consider that DHA consumption is likely to result in increased membrane-associated (lipid-bound) DHA content which is of structural and functional relevance; while phospholipid cleavage under pathological conditions is likely to result in reduced lipid-bound DHA and increased free DHA. In fact, in AD there is a dramatic loss of dendritic spines as well as a significant loss of neurons with a concomitant increase in the levels of astrocytes (reviewed in [
58,
91]). This cell-type remodeling could explain the overall increase in free PUFAs reported here and by others [
82,
105].
Supporting our proposed model in which iPLA
2 activation induces free DHA accumulation, we observed a strong correlation between DHA MALDI-imaging maps and iPLA
2 immunofluorescence. Specifically, we report iPLA
2 immunolabeling within the perinuclear cytoplasm and dendritic arborization of pyramidal neurons. Importantly, these results are in agreement with a previous study that reported high iPLA
2 expression within the hippocampus (i.e. in the nuclear envelope of neurons, dendrites, and axon terminals) and lower expressions within the thalamus and hypothalamus of monkey brains [
87].
On the other hand, AA and cPLA
2 histological studies also revealed an overlap in their localizations. Specifically, both AA and cPLA
2 accumulated within nerve fiber bundles and showed significant levels within thalamic and hypothalamic regions, revealing that cPLA
2 and AA release are highly specific to myelin-rich regions. Previous characterization of cPLA
2 within the rat brain noted high activities and immunoreactivities in the hindbrain, with moderate and low activities/staining in the midbrain and forebrain, respectively [
86]. These results are consistent with high cPLA
2 levels/activities within myelin-rich regions. In fact, recent evidence has revealed a strong cPLA
2 immunoreactivity within axons and oligodendrocytes [
67], further supporting our data/model. Notably, myelin-rich regions besides having a high abundance of phospholipids/sphingolipids, are also highly surrounded by astrocyte processes (where cPLA
2 expression has also been reported [
135]). Taken together, AA release seems to occur preferentially within myelin-rich regions in a process involving oligodendrocyte, astrocyte, and/or axonal cPLA
2.
Our results revealed that all major MAPK pathways are activated by Aβ accumulation (both by APP
WT overexpression and even further by accumulation of mutant Aβ) in the absence of fibrillary amyloid deposits. These results are in agreement with previous reports demonstrating that all major MAPK pathways (i.e., ERK, JNK, and p38 pathways) are activated in vulnerable neurons in patients with AD (reviewed in [
137]). Activated MAPK signaling pathways have been proposed to significantly contribute to AD pathogenesis through various mechanisms including regulation of APP, β- and γ-secretases, and induction of neuronal apoptosis (reviewed in [
57]). Here we are adding cPLA
2 (and potentially iPLA
2 as well) activation to the list of mechanisms by which MAPK mediates AD pathologies. Furthermore, we demonstrated for the first time that soluble oligomeric Aβ is sufficient to dramatically activate MAPK pathways in the absence of amyloid deposition and that this Aβ-induced MAPK activation occurs in a PKC-independent manner.
The importance of wild-type APP controls for mutant APP transgenic studies
Our results revealed that over time overexpression of WT APP is sufficient to induce significant alterations in a broad set of lipid and proteins classes. Although some of these effects were further exacerbated in APP mutant Tg mice comparted to APPWT mice; we found several examples of markers that despite being altered between APP-Tg and non-Tg mice, were not significantly different between APPWT and APP mutant mice (like some PKC isoforms, GAPDH, and β-tubulin). Our results clearly demonstrate that unless APPWT-Tg mice are included as controls, caution needs to be taken when concluding that a given effect is a consequence of a specific APP mutation(s) since such outcome could be partially or fully caused merely by transgenic APP gene expression.
Unfortunately, even though a plethora of mutant APP-Tg mouse models are currently available, only a handful of them have their respective APP
WT-Tg mouse control available. In fact, the vast majority of published AD animal studies have based their conclusions on comparisons between mutant APP-Tg mice versus non-Tg controls. We urge the AD field to consider incorporating APP
WT-Tg control mice into their experimental approaches whenever mutant APP-Tg mice are used. Alternatively, mutant APP-Tg results should be confirmed on human AD brain tissue and/or other non-Tg AD models, like the recently developed APP mutant knock-in mice [
71]. Notably, our results also demonstrated that wild-type APP overexpression on its own is capable of modeling at least some aspects of AD (e.g., MAPK/cPLA
2 activation and free PUFA accumulation). Consistently, overexpression of wild-type hAPP causes early onset familial AD in human carriers with APP duplications [
73,
92] and presumably in Down’s syndrome [
127].