Modulation of Neurolipid Signaling and Specific Lipid Species in the Triple Transgenic Mouse Model of Alzheimer’s Disease

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder and the most common cause of dementia in aging populations. Recently, the regulation of neurolipid-mediated signaling and cerebral lipid species was shown in AD patients. The triple transgenic mouse model (3xTg-AD), harboring βAPPSwe, PS1M146V, and tauP301L transgenes, mimics many critical aspects of AD neuropathology and progressively develops neuropathological markers. Thus, in the present study, 3xTg-AD mice have been used to test the involvement of the neurolipid-based signaling by endocannabinoids (eCB), lysophosphatidic acid (LPA), and sphingosine 1-phosphate (S1P) in relation to the lipid deregulation. [35S]GTPγS autoradiography was used in the presence of specific agonists WIN55,212-2, LPA and CYM5442, to measure the activity mediated by CB1, LPA1, and S1P1 Gi/0 coupled receptors, respectively. Consecutive slides were used to analyze the relative intensities of multiple lipid species by MALDI Mass spectrometry imaging (MSI) with microscopic anatomical resolution. The quantitative analysis of the astrocyte population was performed by immunohistochemistry. CB1 receptor activity was decreased in the amygdala and motor cortex of 3xTg-AD mice, but LPA1 activity was increased in the corpus callosum, motor cortex, hippocampal CA1 area, and striatum. Conversely, S1P1 activity was reduced in hippocampal areas. Moreover, the observed modifications on PC, PA, SM, and PI intensities in different brain areas depend on their fatty acid composition, including decrease of polyunsaturated fatty acid (PUFA) phospholipids and increase of species containing saturated fatty acids (SFA). The regulation of some lipid species in specific brain regions together with the modulation of the eCB, LPA, and S1P signaling in 3xTg-AD mice indicate a neuroprotective adaptation to improve neurotransmission, relieve the myelination dysfunction, and to attenuate astrocyte-mediated neuroinflammation. These results could contribute to identify new therapeutic strategies based on the regulation of the lipid signaling in familial AD patients.


Introduction
The progressive and irreversible deterioration of cognitive functions present in Alzheimer's disease (AD) include a chronic neurodegeneration characterized by pathological hallmarks including the loss of synapses, the intracellular neurofibrillary tangles (NFT) (mostly composed by hyperphosphorylated tau protein) [1], and extracellular neuritic plaques (enriched in Aβ) [2,3]. AD often has comorbidities with other severe human AD brains including several cortical and hippocampal areas [47,48]. CB 1 receptors play a fundamental role in neuroprotection, controlling excitotoxicity events related to neurodegenerative and inflammatory processes [15,49,50]. Regarding lysophospholipid systems, in vivo and in vitro studies have indicated that the neurophysiology of the LPA and S1P is relevant for AD. For example, autotaxin, an enzyme involved in LPA production, with antioxidative properties, is up-regulated in AD patients [51,52]. Moreover, LPA has been implicated in the up-regulation of BACE1 expression [53]. LPA also activates the phosphorylation of GSK-3 enzyme; LPA-mediated activation of GSK-3 occurs in the Rho pathway and may represent an important link between microtubule and microfilament dynamics in AD [54,55]. In addition, LPA has been described as a novel potential mediator in myelination [56] which is disrupted in AD [57]. Finally, the neuroprotective effects of LPA on Aβ 31-35 induced apoptosis in cultured cortical neurons have been reported [16].
The knowledge about the role of S1P-mediated signaling in AD is more recent, but indicates decreased levels of its endogenous ligand in cortical and subcortical areas in brain samples from AD patients [58] which, at least in the cortex, may be the consequence of aberrant functionality or altered expression of the S1P-synthesizing and/or degrading enzymes [59]. The S1P 1 receptors have also been localized in astrocytes [60].
Thus, the present study analyzed the activities mediated by CB 1 , LPA 1 , and S1P 1 receptors, respectively, in 6-month-old 3xTg-AD transgenic mice using the functional [ 35 S]GTPγS autoradiography method. The receptor specificity was localized at the subcellular level by microscopic immunofluorescent localization. These results are compared to the anatomical distribution patterns of the lipid distribution in the brain of 3xTg-AD mice, obtained by MALDI mass spectrometry imaging (MALDI-MSI).

[ 35 S]GTPγS Binding Assay in 3xTg-AD Mice Brain Sections
In order to assess changes between 3xTg-AD mice and age-matched wild type (WT) mice, firstly, we examined the basal G protein activity in whole brain, which was quite similar in 3xTg-AD and WT mice. The functional coupling induced by WIN55,212-2, accounting mainly for CB 1 -mediated activity, was decreased in transgenic mice compared to WT mice in the amygdala (3xTg-AD 112.8 ± 28.9% vs. WT 295.5 ± 41.7%; p < 0.01). This CB 1 activity was also decreased in the VI layer of the motor cortex (3xTg-AD 238.4 ± 22.9% vs. WT 435.4 ± 58.2%; p < 0.01) ( Figure 1, Table 1).
On the other hand, the [ 35 S]GTPγS binding induced by LPA was increased in transgenic mice at the striatum (3xTg-AD 23.1 ± 3.8% vs. WT 3.3 ± 7.2%; p < 0.05), motor cortex (3xTg-AD 26.1 ± 7.6% vs. WT 6.2 ± 12.4%; p < 0.05), corpus callosum (3xTg-AD 189.6 ± 17.4% vs. WT 90.8 ± 12.3%; p < 0.05), and hippocampal CA1 area (3xTg-AD 22.7 ± 4.4% vs. WT −18.7 ± 7.8%; p < 0.05) ( Figure 2, Table 2). demonstrated in frontal cortex samples in AD patients [45]. Furthermore, CB1 receptors are up-regulated in the early stages of AD, followed by a diminution of its availability during the progress of the disease [40,46]. In contrast, other studies found no alteration of CB1 densities in AD brains including several cortical and hippocampal areas [47,48]. CB1 receptors play a fundamental role in neuroprotection, controlling excitotoxicity events related to neurodegenerative and inflammatory processes [15,49,50]. Regarding lysophospholipid systems, in vivo and in vitro studies have indicated that the neurophysiology of the LPA and S1P is relevant for AD. For example, autotaxin, an enzyme involved in LPA production, with antioxidative properties, is up-regulated in AD patients [51,52]. Moreover, LPA has been implicated in the up-regulation of BACE1 expression [53]. LPA also activates the phosphorylation of GSK-3 enzyme; LPA-mediated activation of GSK-3 occurs in the Rho pathway and may represent an important link between microtubule and microfilament dynamics in AD [54,55]. In addition, LPA has been described as a novel potential mediator in myelination [56] which is disrupted in AD [57]. Finally, the neuroprotective effects of LPA on Aβ 31-35 induced apoptosis in cultured cortical neurons have been reported [16]. The knowledge about the role of S1P-mediated signaling in AD is more recent, but indicates decreased levels of its endogenous ligand in cortical and subcortical areas in brain samples from AD patients [58] which, at least in the cortex, may be the consequence of aberrant functionality or altered expression of the S1P-synthesizing and/or degrading enzymes [59]. The S1P1 receptors have also been localized in astrocytes [60].
Thus, the present study analyzed the activities mediated by CB1, LPA1, and S1P1 receptors, respectively, in 6-month-old 3xTg-AD transgenic mice using the functional [ 35 S]GTPγS autoradiography method. The receptor specificity was localized at the subcellular level by microscopic immunofluorescent localization. These results are compared to the anatomical distribution patterns of the lipid distribution in the brain of 3xTg-AD mice, obtained by MALDI mass spectrometry imaging (MALDI-MSI).

[ 35 S]GTPγS Binding Assay in 3xTg-AD Mice Brain Sections
In order to assess changes between 3xTg-AD mice and age-matched wild type (WT) mice, firstly, we examined the basal G protein activity in whole brain, which was quite similar in 3xTg-AD and WT mice. The functional coupling induced by WIN55,212-2, accounting mainly for CB1-mediated activity, was decreased in transgenic mice compared to WT mice in the amygdala (3xTg-AD 112.8 ± 28.9% vs. WT 295.5 ± 41.7%; p < 0.01). This CB1 activity was also decreased in the VI layer of the motor cortex (3xTg-AD 238.4 ± 22.9% vs. WT 435.4 ± 58.2%; p < 0.01) ( Figure 1, Table 1).   The p values were calculated by two-tailed Student's t test ** p ≤ 0.01.

GPCR-Immunoreactivity and Astrocyte Density
The immunosignaling associated to CB 1 and S1P 1 receptors was observed in the gray matter, whereas that associated to LPA 1 receptor was mainly restricted to white matter regions and only modestly to discrete regions of the gray matter ( Figure 4). These observations are in accordance with the observed distribution of the different GPCR functional activity in the autoradiographic studies. In 3xTg-AD mice, S1P 1 immunoreactivity was found to be decreased in the hippocampus, and LPA 1 was increased in the corpus callosum, which is in accordance with that observed in the functional coupling to G i/o proteins evoked by the different agonists. Collectively, autoradiographic and immunohistochemical results demonstrate that the changes in the functional activity of these receptors may be directly related to intrinsic variations in the density of these receptors in 3xTg-AD mice.
Some of these three neurolipid receptors have also been localized in astrocytes, mainly the S1P 1 receptor subtype. Therefore, the study of astroglial cells by immunofluorescence was performed in those brain regions which showed marked differences in the functional coupling of the analyzed GPCRs. Two different markers, glial fibrillary acidic protein (GFAP) and S100B, were used to identify astrocytes directly in the tissue. The immunosignal observed by using both markers exhibited different immunostaining patterns showing that astrocytes were mainly distributed in the gray matter. GFAP immunoreactivity clearly delineated the body and the processes of the astrocytes, whereas S100B immunosignal was more restricted and mainly confined to the astrocyte body. The total density of Hoechststained nuclei was not modified in transgenic mice. In this sense, both the density of astrocytes and the total area stained with GFAP or S100B were normalized as percentages of total nuclei or total Hoechst-stained area, respectively. Marked changes in density, as well as in cell size, were found in hippocampal CA1 and dentate gyrus fields in 3xTg-AD mice, demonstrating not only a decrease in the population of astrocytes, but also their atrophy or shrinkage. These observations allowed us to clearly differentiate between both genotypes depending on the astroglial-associated immunosignal. However, those differences were not statistically significant in other brain regions such as the granular olfactory bulb ( Figure 5). tional activity in the autoradiographic studies. In 3xTg-AD mice, S1P1 immunoreactivity was found to be decreased in the hippocampus, and LPA1 was increased in the corpus callosum, which is in accordance with that observed in the functional coupling to Gi/o proteins evoked by the different agonists. Collectively, autoradiographic and immunohistochemical results demonstrate that the changes in the functional activity of these receptors may be directly related to intrinsic variations in the density of these receptors in 3xTg-AD mice. Immunolabeling of CB1 (i, green), LPA1 (ii and iii, red), S1P1 (iv, red) receptors in the hippocampus (i, ii and iv) and corpus callosum (iii) from WT (A) and 3xTg-AD mice (B). Note that CB1 and S1P1 receptors are distributed in the gray matter, whereas LPA1 receptors are mainly expressed in white matter regions. The different hippocampal subfields exhibit specific immunostaining patterns. CA1; CA1 region of hippocampus, DG; dentate gyrus of hippocampus, Or; oriens layer of CA1, Py; pyramidal layer of CA1, Rad; radiatum layer of CA1, Mol; molecular layer of DG, GrDG; granular layer of DG, Wm; white matter, cc; corpus callosum, Cing cx; cingular cortex. Scale bar = 200 µm.
Some of these three neurolipid receptors have also been localized in astrocytes, mainly the S1P1 receptor subtype. Therefore, the study of astroglial cells by immunofluorescence was performed in those brain regions which showed marked differences in the functional coupling of the analyzed GPCRs. Two different markers, glial fibrillary acidic protein (GFAP) and S100B, were used to identify astrocytes directly in the tissue. The immunosignal observed by using both markers exhibited different immunostaining patterns showing that astrocytes were mainly distributed in the gray matter. GFAP im- . Immunolabeling of CB 1 (i, green), LPA 1 (ii and iii, red), S1P 1 (iv, red) receptors in the hippocampus (i, ii and iv) and corpus callosum (iii) from WT (A) and 3xTg-AD mice (B). Note that CB 1 and S1P 1 receptors are distributed in the gray matter, whereas LPA 1 receptors are mainly expressed in white matter regions. The different hippocampal subfields exhibit specific immunostaining patterns. CA1; CA1 region of hippocampus, DG; dentate gyrus of hippocampus, Or; oriens layer of CA1, Py; pyramidal layer of CA1, Rad; radiatum layer of CA1, Mol; molecular layer of DG, GrDG; granular layer of DG, Wm; white matter, cc; corpus callosum, Cing cx; cingular cortex. Scale bar = 200 µm.

Figure 5.
Quantitative analyses of the astrocyte population according to the following; number of cells immunolabeled with GFAP (A) or S100B (B) over the total cell nuclei (i), total area immunolabeled with GFAP or S100B over the total area occupied by cell nuclei (ii). Quantitative analyses of the immunolabeled area with GFAP or S100B of individual cells show the astrocyte size (µm 2 ) in both genotypes (iii). p < 0.05 vs. WT mice. CA1; hippocampus CA1, DG; hippocampus dentate gyrus, OB; olfactory bulb, including granular and anterior. Hoechst staining of nuclei and double labeling of astrocytes in brain tissue from WT and 3xTg-AD revealing particular immunostaining patterns observed with GFAP and S100B, which stain astrocytic processes or cell bodies, respectively (Ci and Cii). Note the marked decrease in the density of astrocytes as well as their atrophy in the transgenic genotype (3xTg-AD). Cii scale bar = 40 µm.
The more significant differences in negative ion detection mode were for the following lipid species: SM  Figure 6). Quantitative analyses of the astrocyte population according to the following; number of cells immunolabeled with GFAP (A) or S100B (B) over the total cell nuclei (i), total area immunolabeled with GFAP or S100B over the total area occupied by cell nuclei (ii). Quantitative analyses of the immunolabeled area with GFAP or S100B of individual cells show the astrocyte size (µm 2 ) in both genotypes (iii). * p < 0.05 vs. WT mice. CA1; hippocampus CA1, DG; hippocampus dentate gyrus, OB; olfactory bulb, including granular and anterior. Hoechst staining of nuclei and double labeling of astrocytes in brain tissue from WT and 3xTg-AD revealing particular immunostaining patterns observed with GFAP and S100B, which stain astrocytic processes or cell bodies, respectively (Ci and Cii). Note the marked decrease in the density of astrocytes as well as their atrophy in the transgenic genotype (3xTg-AD). Cii scale bar = 40 µm.

MALDI-MSI Assay in 3xTg-AD Mice Brain Sections
The most significant differences in positive ion detection mode between 3xTg-AD and WT mice were found in the following lipid species: PA[(34:1) + K] + ; cortex (3xTg-AD 25.0 ± 2.6% vs. WT Figure 6). Table 4. Percentage of the intensity of molecular lipid species in positive and negative mode in sagittal mice sections from WT (n = 6) compared to the 3xTg-AD (n = 6), as revealed by MALDI-MSI. Data are mean ± SEM values.

Discussion
The 3xTg-AD mouse model of AD is an experimental animal model that has been employed for the examination and evaluation of the effects during the development of some of the mechanisms that have been related to genetic familial forms of AD. The aim of the present study was to analyze together the activity of the main receptors for neurolipids present in the central nervous system (CNS): CB1, LPA1, and S1P1 in 3xTg-AD mice at 6 months. The results are discussed in the framework of the lipid composition of the brains in these mice obtained by the MALDI-MSI technique for the in situ analysis, contributing to understand the lipid changes already observed in AD patients and connect The more significant differences in negative ion detection mode were for the following lipid species: SM  Figure 6).

Discussion
The 3xTg-AD mouse model of AD is an experimental animal model that has been employed for the examination and evaluation of the effects during the development of some of the mechanisms that have been related to genetic familial forms of AD. The aim of the present study was to analyze together the activity of the main receptors for neurolipids present in the central nervous system (CNS): CB 1 , LPA 1 , and S1P 1 in 3xTg-AD mice at 6 months. The results are discussed in the framework of the lipid composition of the brains in these mice obtained by the MALDI-MSI technique for the in situ analysis, contributing to understand the lipid changes already observed in AD patients and connect these results with the possible adaptations in the activity induced by three different neurolipid-mediated signaling systems: eCB, LPA, and S1P.

Modulation of CB 1 Receptor Activity
Concerning the cannabinoid system, the activity of CB 1 receptors, measured as WIN55,212-2-induced [ 35 S]GTPγS binding, was lower in the posterior amygdala and layer VI of the motor cortex of 3xTg-AD mice when compared with age-matched WT animals. The cerebral cortex and amygdala belong to the CNS emotional circuitry and contain high levels of CB 1 receptors [61]. It has been described that the cannabinoid signaling in the prefrontal cortex can modulate the magnitude of neuronal emotional learning plasticity and memory formation through functional inputs from the basolateral amygdala [62]. The amygdala is a region of the temporal lobe that is affected by Aβ and neurofibrillary tangle pathology at early stages of AD. In 3xTg-AD mice, an increase of anxiety and fear related behaviors has been observed and, at the time when Aβ is still localized intraneuronally, some spatial memory deficits appear [63]. In 3xTg-AD mice, Aβ accumulation occurs preferably inside the amygdaloid glutamatergic neurons, where CB 1 receptors are also located.
Furthermore, we evaluated if the availability and distribution of the CB 1 receptors in 3xTg-AD mice could account for the above-described results. The analysis of the [ 3 H]CP55,940 binding sites in 3xTg-AD mice and age-matched control animals revealed a significant increase of CB 1 receptor densities in different areas of the 3xTg-AD mice such as the substantia nigra, cerebellum gray matter, dorsal hippocampal CA1 area, and cingular cortex. Studies of CB 1 receptors in 6-month-old and 10-month-old AβPP/PS1 mice have shown a decrease in CB 1 receptors in the cortex and hippocampus, respectively [64,65]. Moreover, AβPP/PS1 mice presented higher levels of CB 1 receptor in the cortex than wild-type mice at 3 months of age [66,67]. Recent studies have reported high levels of CB 1 mRNA and functional protein in 6-month-old and 7-month-old 3xTg-AD mouse brain in the prefrontal cortex, dorsal hippocampus, and basolateral amygdala [68,69]. The data obtained from the [ 3 H]CP55,940 autoradiography did not correlate with the [ 35 S]GTPγS binding stimulated by WIN55,212-2, suggesting that receptor density and receptor efficiency can be modulated separately and the contribution of CB 1 receptors coupled to G q proteins could account for these discrepancies [70]. Previous studies based on human postmortem brain samples suggested that CB 1 receptors could be involved in the pathophysiology of AD [45,46,48,71]. Our research group has observed in patients an increase in CB 1 density at layer VI of the frontal cortex and different areas of the hippocampus, such as pyramidal layer during the moderate stages of the disease, but having a significant decrease at later stages in the pyramidal layers of the different hippocampal areas and the inner layers of the entorhinal cortex [40].
The decrease of the CB 1 signaling in 3xTg-AD mice was detected in brain areas innervated by basal forebrain cholinergic neurons, the posterior amygdala, and inner layers of the motor cortex; therefore, a modulation in this pathway on demand of lipid precursors for the further synthesis of eCB could be expected. The synthesis of eCB starts with the release of cell membrane phospholipid precursors such as phosphatidylcholines to further obtain N-arachidonoyl phosphatidylethanolamine or, phosphatidylinositols to further obtain diacylglycerols, in order to synthesize either anandamide or 2-AG, respectively [72]. Interestingly, at least three lipid species which may contain an arachidonic acid (AA) (20:4) moiety were found to be consistently decreased throughout different brain regions including the cortex, the hippocampus, the amygdala, and the cerebellum in the 3xTg-AD mice. MALDI-MSI analyses showed that certain phospholipid species such as PI(16:0/20:4) − and PC(38:6) + which are decreased in the cortex and in the hippocampus, as well as PC(36:4) + species, which is decreased in the amygdala and in the cerebellum, may be being exploited from any membrane pool for this precise biosynthetic process as a physiological adaptation for the observed dysregulation of the CB 1 -mediated signaling.

Modulation of LPA 1 Receptor Activity
In the present study, an increase in LPA 1 activity (LPA induced [ 35 S]GTPγS binding) was observed in the corpus callosum, motor cortex, hippocampal CA1 area, and striatum of 3xTg-AD mice. Several studies have described that LPA 1 receptors are expressed in most cell types of the CNS, including neuronal progenitors [73], astrocytes [74], microglia [75,76], and oligodendrocytes [17,77]. We have shown a decrease of microglia and astrocytes in CA1 and DG hippocampal areas of triple transgenic mice, suggesting the increase in LPA 1 receptor activity in this area could be a compensatory effect due to the decrease of microglia and astrocytes [45]. However, there are other factors that could be related to that increase in LPA 1 receptor activity, such as myelination disruption. It has been described some degree of myelination disruption in 6-month-old 3xTg-AD mice (early pathological stage) in subregions of hippocampus and entorhinal cortex, together with hyperphosphorylated tau, and a decline of myelin basic protein and 2',3'-Cyclic-nucleotide 3'-phosphodiesterase expression levels, which are myelin and oligodendrocytes major proteins [78]. Furthermore, in 12-month-old APP Swe mice, myelination defects have also been described in the corpus callosum [79]. Oligodendrocyte myelin sheath integrity is necessary for axon viability and for the maintenance of axonal flow [80]. LPA 1 receptor has been reported as a novel marker for differentiated oligodendrocytes, suggesting that initiation of LPA 1 expression may contribute to the myelinating oligodendrocyte phenotype [56,81]. The increase of LPA 1 receptor activity that we observed at 6-month-old mice (initial stage of the disease) might indicate a neuroprotective action mediated by LPA in response to initial white matter damage. White matter dysfunction seems to appear prior to amyloid or tau pathology in different AD mice models [78,82]. Myelination processes are a vulnerable target contributing to early disease progression. Furthermore, the increased LPA 1 receptor activity in 3xTg-AD mice at cortex and striatum coincides with the significant increase of PA(34:1) lipid species in the same areas. This increase could be associated with LPA production since LPA is generated on demand from PA by the phospholipase A 2 enzyme. Interestingly, LPA 18:1 is the most abundant LPA species in brain [83] and PA(34:1) is constituted by oleic acid (18:1). Therefore, the increased LPA 1 receptor activity could yield to adaptations during the development of 3xTg-AD mice, increasing the demand of LPA endogenous neurotransmitter and increasing the levels of lipid precursors such as PA(34:1) [84].

Modulation of S1P 1 Receptor Activity
The subtype 1 of sphingosine-phosphate receptors (S1P 1 ), a lysophospholipid G i/ocoupled GPCR, which is activated by the endogenous neurolipid S1P, was also analyzed. We found an intense S1P 1 activity in the CNS of both genotypes, but marked reductions in the functional coupling to G i/o proteins in the transgenic mice following the activation with the specific S1P 1 agonist CYM5442. The S1P 1 activity was mainly restricted to gray matter, and was even higher than that observed for CB 1 receptor activity in several brain regions. This S1P 1 activity may be related to the modulation of neuroinflammatory processes [85]. S1P 1 receptors are highly expressed in astrocytes and the loss of hippocampal S1P 1 -mediated signaling could be explained due to the loss and/or atrophic processes on astrocytes. This phenomenon had previously been described in the entorhinal cortex from 3xTg-AD mice, and explained as the loss of astrocyte-mediated anti-inflammatory response to Aβ accumulation [86]. Accordingly, the present study shows a clear reduction of S1P 1 activity in hippocampal areas and a tendency to decrease in the entorhinal cortex from 3xTg-AD mice. The study of astrocytes carried out in the present work using double immunofluorescence images of GFAP/S100B markers clearly demonstrated that both dies are useful to detect and quantify astrocytes. Moreover, the images show particular immunoreactivity patterns that confirm both the decrease of the astrocyte population and their atrophy in the hippocampus. Together, these results are consistent with the idea that following a reduction in the astrocyte-mediated response against the Aβ accumulation, decreased S1P 1 -mediated signaling could contributes to reduce neuroinflammatory responses in this AD mice model. The role of S1P 1 -mediated signaling in AD remains poorly understood, however, the lower expression of sphingosine kinase-1 (S1P-synthesizing enzyme) together with enhanced expression of S1P lyase (S1P-degrading enzyme) lead to the loss of the S1P endogenous ligand pool early in AD [58,59]. Since the use of drugs targeting S1P signaling such as fingolimod (Gilenya ® , Novartis Pharma AG, Basel, Switzerland), which induces a functional antagonism of S1P 1 receptors (i.e., reducing S1P 1 -mediated signaling), was approved as immunotherapeutic drug for the treatment of multiple sclerosis, further studies in AD models will contribute to explore the potential of S1P 1 agonists also for AD treatment. In this sense, the administration of fingolimod to a mouse model overexpressing Aβ led to improve Aβ-associated pathology by attenuating the neuroinflammatory response [85]. It is not clear if the general decrease of S1P 1 signaling observed in 6-month-old 3xTg-AD mice is a compensatory mechanism to counteract neuroinflammatory events or conversely, is contributing to worsen the pathology, but these evidences point to this neurolipid signaling system as a promising pharmacological target for the treatment of neurodegenerative diseases.
The SM lipid species represent the main phospholipid pool to further synthesize S1P. The increase on the relative abundance of two particular species, SM(18:1/18:0) and SM(d35:1) observed in the hippocampus, where the activity mediated by S1P 1 receptor was found to be decreased, may indicate a possible cause-effect relationship. The hypoactivity of S1P 1 -mediated signaling in triple transgenic mice could result in the accumulation of these particular SM species due to a lower requirement in the synthesis of S1P. On the other hand, ceramides are well known intermediates in the metabolic pathways of sphingolipids and one would expect to find a decrease in their levels. These ceramides were not detected by MALDI-MSI analysis under the present experimental conditions, preventing a more complete analysis of the metabolic turnover of sphingolipids in this model of AD. These sphingolipids are implicated in the programmed cell death and are directly involved in neurodegeneration, particularly in AD [87]. Interestingly, increased levels of SM(d18:1/18:0) have been found in the hippocampal gray matter as well as in cerebrospinal fluid from AD patients [88,89]. Recently, a relation between ceramide generation and a reduction in mitochondrial ATP release has been reported in astrocytes [90]. In agreement with the present immunofluorescence study, increased levels of specific SM species could be involved in the observed changes in astroglial density and size in 3xTg-AD mice and in relation with S1P signaling. Nevertheless, additional correlational studies measuring the endogenous levels of ceramides and S1P, the enzymatic machinery associated to sphingolipid metabolism, as well as S1P 1 receptor density will contribute to clarify this issue.

Anatomical Localization of Lipid Species in 3xTg-AD Mice Brain by MALDI-MSI Assay
The field of neurolipidomics tries to understand how dynamic changes in membrane lipid composition are contributing to regulate brain cell function. Previous studies have indicated that lipid molecules play a relevant role in AD, and some of these lipids have frequently been reported at abnormal concentrations in AD tissue [35,[91][92][93]. Although several studies have been performed with AD transgenic models showing lipid impairments [94][95][96][97][98], the present study is pioneer in achieving anatomical localization of lipid species in 3xTg-AD mice brain by MALDI-MSI assay. We have observed modifications on PC, PA, SM, and PI intensities in different brain areas. Moreover, the modulations of PC and PI species depend on their fatty acid composition, i.e., decrease of polyunsaturated fatty acid (PUFA) phospholipids and increase of phospholipid species containing saturated fatty acids (SFA). Similar results have been reported in serum and tissue of AD patients [31][32][33]35] and also in another AD transgenic mice model [97,99]. The decrease in phospholipids containing PUFA could be related to impairment of the cell membranes during AD pathogenesis. Some morphological and neurofunctional damages have been found to correlate with PUFA declines, including swollen astrocytes, deformed nerve cell nuclei, reduced acetylcholine release, and modifications on the fluidity, structure, and permeability of the cell membranes [100][101][102]. Furthermore, the increase in the PA(34:1) species in 3xTg-AD mice brain, could be related with an increase in total phospholipase D activity that has been reported in AD brain homogenates, using an in vitro enzymatic assay [103]. In addition, decrease in PC species have been described as possible plasma biomarkers for AD, even in patients before the onset of the disease [104]. In our study, we have found a decrease in one of that PC species, that could be induced by upregulation of PLA 2 enzyme in AD [105,106]. In contrast, we have not found a decrease of glucosylceramides or sulfatides as have been reported in the APP/PS1 and APP/tau transgenic mice lineages or even in patients at the first stages of the disease [97,98,107].
In summary, the modulation of the main CNS receptors of the LPA, eCB, and S1P neurolipid systems analyzed in the triple transgenic model of AD suggests a neuroprotective adaptation during the development of these mice. The cannabinoid activity improving or maintaining the neurotransmission, LPA activity trying to relieve the myelination dysfunction in the axons, and S1P 1 activity attenuating astrocyte-mediated neuroinflammatory response [108]. In addition, the observed changes on lipid species in the 3xTg-AD mice in specific brain regions suggest a similar modulation in the cases of familial AD patients, which are covered by this mouse model.
Further studies will help us to shed light on the relevance of the observed modifications and if they are indicating primary effects or are a physiological outcome of the neurodegeneration. The complexity of the AD biochemistry in the brain is probably a consequence of multiple causes that are converging in the observed clinical manifestations that include the progressive dementia. The research on neurolipid signaling and their control on the lipid homeostasis and modulation of other neurotransmitter systems has been limited by the techniques used to anatomically identify the super-specialization on lipid species in the brain, which reaches the highest levels in the human cortex. The MSI used in the present study combined with other neuroanatomical methods will open new perspectives in our ultimate goal of understanding the integration of energetic, structural, and signaling functions mediated by lipid molecules in the brain that will contribute to develop specific and effective treatments for neuropsychiatric and neurodegenerative diseases, including AD.

Animals and Tissue Preparation
Triple transgenic mice (3xTg-AD) were obtained from Department of Psychiatry and Forensic Medicine, Universitat Autònoma de Barcelona, Barcelona, Spain, in collaboration with Dr. Lydia Giménez-Llort. 3xTg-AD mice harboring PS1 M146V , APP Swe and tau P301L transgenes were genetically engineered at the University of California Irvine, as previously described [19]. Briefly, two independent transgenes (encoding human APP Swe and human tau P301L , both under control of the mouse Thy1.2 regulatory element) were co-injected into single-cell embryos harvested from homozygous mutant PS1 M146 V knock-in (PS1KI) mice.
Six-month-old male 3xTg-AD mice (n = 26) and WT mice with the same background but without genetic modifications (n = 22) were used. The breeding program was established at the Universitat Autònoma de Barcelona. All the animals were housed and maintained under standard laboratory conditions (12 h light:dark, cycle starting light at 8:00 am, food and water available ad libitum, 22 ± 2 • C, 50-60% humidity). Animals were transferred to the animal department of UPV/EHU, with the same standard housing conditions, one month before the experimental procedures.

Tissue Preparation
Mice were deeply anesthetized with ketamine/xylazine (90/10 mg kg −1 ; i.p.). Fresh tissue. The brain samples were quickly removed by dissection, fresh frozen, and kept at −80 • C. Later, the brains were cut on a Microm HM550 cryostat (Thermo Fisher Scientific, Whaltham, MA, USA) to obtain 20 µm sections that were mounted onto gelatin-coated slides and these were stored at −20 • C until used.

[ 35 S]GTPγS Binding Assay
The tissue sections were air-dried for 15 min. Then, slides containing the sections were washed in a HEPES based buffer containing 50 mM HEPES, 100 mM NaCl, 3 mM MgCl 2 and 0.2 mM EGTA, 0.5% bovine serum albumin (BSA) at pH 7.4, for 30 min at 30 • C in a water bath. The pre-incubation was repeated a second time in new buffer to ensure the washing of endogenous GPCR ligands. In a second step, the slides were incubated for 2 h at 30 • C in a solution containing 2 mM guanosine diphosphate (GDP), 1 mM DL-dithiothreitol (DTT), adenosine deaminase (3 u/L) and 0.04 nM [ 35 S]GTPγS. The agonist-stimulated binding was measured under the same conditions but in the presence of the specific GPCR agonists: LPA (10 −5 M), WIN55,212-2 (10 −5 M) and CYM5442 (10 −5 M). Ki16425 (10 −5 M) was used together with LPA, AM251 (10 −5 M) with WIN55,212-2, and W146 (10 −5 M) with CYM5442 as respective antagonists to validate that the assays were specific of the receptor subtype. Non-specific binding was determined in the presence of 10 µM of non-labelled GTPγS. Sections were washed twice in an ice-cold HEPES buffer 50 mM (pH 7.4), dipped in distilled water, and air-dried. Sections were exposed to Kodak Biomax MR films (Sigma, St. Louis, MO, USA) together with 14 C standards for 48 h at 4 • C.

Quantitative Image Analysis of Film Autoradiograms
Films were scanned and quantified by transforming the optical densities into nCi/g tissue equivalent (nCi/g t.e.) and percentage of stimulation over the basal (%) was calculated using an image analysis system (NIH-IMAGE, Bethesda, MA, USA). (U.S National Institutes of Health, http://rsb.info.nih.gov/nih-image/). This software defines the optical density of an anatomical area from 0 (white) to 256 (black). The [ 14 C] radioactive standards that were co-exposed with the slides were used to calibrate the optical densities with the level of radioactivity labeled to the sections. Experimental data were analyzed by using the computer programs GraphPad Prism (v. 5.0, Graph Pad) and Microsoft office Excel 2007. Data were expressed as the mean values ± SEM. Differences between regions were analyzed by unpaired two-tailed Student's t test.

Quantitative Analyses of Astrocytes
Sections were inspected and immunofluorescence images from WT (n = 3) and 3xTg-AD (n = 3) were used to quantify the astrocyte density; 400-fold magnification photomicrographs (SPOT Flex Shifting Pixel CCD imaging camera) were acquired on an Axioskop 2 Plus epifluorescence microscope (Carl Zeiss, Oberkochen, Germany) in both hemispheres under the same microscopic conditions. Using Image J software (NIH, Bethesda, MD, USA), images were converted to a binary mode and different processes were performed in order to identify single astrocytes and nuclei by applying the watershed option. The total number of astrocytes and nuclei (N/mm 2 ) were quantified, and the total area (in pixels) stained by astrocytes (S100B + or GFAP + -immunoreactivity) or nuclei (Hoechst staining) and each astrocyte and nuclei stained area (size in pixels) were calculated in each image. Hoechst stained nuclei were used to normalize the number of GFAP or S100B positive cells in each image (% of astrocytes of total nuclei). Hoechst stained area was used to normalize the GFAP and/or S100B positive area in each image (% of GFAP or S100B immunopositive area of total Hoechst-stained area).

Sample Preparation for MALDI-MSI
The original lipid composition and anatomical characteristics of the tissue must be preserved throughout the sample-preparation process [109]. The brains were cut on a Microm HM550 cryostat to obtain 20-µm sections and stored at −20 • C until the moment of use.
Once the initial tissue preparation steps had been completed, the chemical matrix was deposited on the tissue surface prior to analysis by sublimation. For tissue sections mounted on glass slides, sublimation was performed using 300 mg of mercaptobenzothiazole (MBT), by controlling the deposition time and temperature (30 min at 140 • C), making it possible to control the thickness of the matrix layer and optimize the s/n ratio of the mass spectra, avoiding lipid migration thanks to the lack of solvent. Finally, a re-crystallization of the sample was performed, using a normal glass Petri plate (100 mm diameter × 15 mm depth (Thermo Fisher Scientific, Whaltham, MA, USA)) as following. Thus, 1 mL of methanol (99%) was deposited onto a piece of paper previously placed in the bottom of the Petri plate in order to create a vapor atmosphere for the re-crystallization process on a hot plate (1 min at 40 • C). This step allowed us to achieve a higher intensity in the detection of the peaks [110].

Mass Spectrometer
A MALDI LTQ-XL-Orbitrap (Thermo Fisher, San Jose, CA) equipped with a nitrogen laser (λ = 337 nm, rep. rate = 60 Hz, elliptical spot size = 80 × 120 µm 2 ) was used for mass analysis. Thermo's ImageQuest TM 1.0.1 and Xcalibur TM 3.1 software were used for MALDI-MSI data acquisition. The images were acquired in both negative and positive ion mode. The positive ion range was 500-1000 Da, with 10 laser shots per pixel at a laser fluence of 15 µJ. The negative ion range was 400-1100 Da, with 10 laser shots per pixel at laser fluence of 15 µJ. The target plate stepping distance was set to 150 µm for both the xand y-axes by the MSI image acquisition software. The mass resolution was 100,000 in both positive and negative ion mode. The data were normalized using the total ion current to avoid the displacement in masses along the tissue caused by irregularities on the surface or other experimental artifacts.

Image and Spectra Analysis for MALDI-MSI
The MALDI images were generated using the ImageQuest software (Thermo Scientific, San Jose, CA, USA). With this software, a m/z range is plotted for signal intensity for each pixel (mass spectrum) across a given area (tissue section). The quality of the images was improved during the image creation process by selecting the m/z range of interest and doing a normalization as a ratio of total ion current (TIC) for each mass spectrum. Different regions of interest (ROI) were analyzed including hippocampus, cortex, amygdala, cerebellum, and striatum. The spectra intensity was further normalized as a ratio of the peak or m/z value with the highest intensity, PC[(34 + 1) + K] + in positive ion mode and PI[18:0/20:4] − in negative ion mode and the average was calculated using the OriginPro 8 software. The most intense peak was considered the 100% and the intensity of the rest of the peaks was calculated as a percentage. The two-tailed unpaired Student's t-test was used for the comparison of two groups. The results were considered significant when p ≤ 0.05.

Peak Assignment
The assignment of the m/z values to specific molecules is complex in this type of studies, usually containing a large number of lipids that share similar masses. Therefore, the assignment of lipid species was facilitated using databases such as Lipid MAPS (http://www.lipidmaps.org/ accessed on 9 November 2021), and different reported articles. A 5 ppm mass accuracy was used as the tolerance window. The glycerolipid species numbers (x:y) denote the total lengths and the number of double bonds of the acyl chains, while the sphingolipid and sulfatide species numbers correspond to the length and number of double bonds of the acyl chain added to those of the attached sphing-4-enine (d18:1) or sphinganine (d18:0) base.