ABSTRACT
INTRODUCTION ApoE is a carrier for brain lipids and the most important genetic risk factor for Alzheimer’s disease (AD). ApoE binds the receptor sortilin which mediates uptake of apoE-bound cargo into neurons. The significance of this uptake route for brain lipid homeostasis and AD risk seen with apoE4, but not apoE3, remains unresolved.
METHODS Combining neurolipidomics in patient specimens with functional studies in mouse models, we interrogated apoE isoform-specific functions for sortilin in brain lipid metabolism and AD.
RESULTS Sortilin directs uptake and conversion of polyunsaturated fatty acids into endocannabinoids, lipid-based neurotransmitters that act through nuclear receptors to sustain neuroprotective gene expression in the brain. This sortilin function requires apoE3, but is disrupted by binding of apoE4, impairing endocannabinoid signaling and increasing amyloidogenic processing.
DISCUSSION We uncovered the significance of neuronal apoE receptor sortilin in facilitating neuroprotective actions of brain lipids, and its relevance for AD risk seen with apoE4.
BACKGROUND
Apolipoprotein (apo) E is the major carrier for lipids in the brain. It is secreted by astrocytes and microglia and delivers essential lipids to neurons that take up apoE-bound cargo via apoE receptors (reviewed in [1]). Apart from its role in lipid homeostasis, apoE also bears significance as the most important genetic risk factor for the sporadic form of Alzheimer’s disease (AD) as carriers of the APOE4 gene variant are at a significantly higher risk of AD than carriers of the common APOE3 allele [2]. A large body of work confirmed the potential of apoE4, as compared with apoE3, to promote brain deposition of amyloid-β peptides (Aβ), a major cause of neurodegeneration (reviewed in [3]). Many hypotheses have been advanced how apoE may affect brain Aβ levels, including the delivery of bound Aβ to cellular catabolism via apoE receptors. Still, the mechanism(s) that distinguish apoE3 and apoE4 functions in brain lipid homeostasis and progression of AD remain controversial.
Previously, we identified the lipoprotein receptor sortilin as a major endocytic route for uptake of apoE-containing lipoproteins in neurons in vitro and in vivo. In gene-targeted mice, loss of sortilin impaired neuronal clearance of murine apoE and was associated with enhanced accumulation of Aβ peptides and senile plaque formation [4]. However, the relevance of the sortilin-dependent uptake of apoE for brain lipid homeostasis and for the risk of AD seen in carriers of the human APOE4 genotype remained unclear.
Combining MS-based lipidomics in patient specimens with functional studies in humanized mouse models expressing apoE3 or apoE4, we uncovered a unique role for sortilin and apoE3 in facilitating the neuronal metabolism of polyunsaturated fatty acids (PUFA) into endocannabinoids (eCB) that signal an anti-inflammatory gene expression profile in the brain. The ability of sortilin to sustain neuroprotective eCB signaling is disrupted by binding of apoE4, increasing pro-inflammatory markers and aggravating the amyloidogenic burden in the brain.
METHODS
Materials and general methods
AD specimens were collected from donors from whom written informed consent for the use of the material for research purposes had been obtained (see supplementary methods). Mouse strains and details on animal experimentation are also given in supplementary methods. Determination of transcript or protein levels in tissue were performed by qRT-PCR and SDS-PAGE, respectively, using standard protocols. Cell culture experiments involving CHO cells are detailed in supplementary methods.
Lipid analyses
Quantification of levels of total cholesterol, triglycerides, and free fatty acids in plasma was performed using commercial kits (Biovision, Roche, Cayman). Mouse plasma lipoprotein profiles were established by TNO Biosciences (Leiden, The Netherlands) using FPLC. Targeted lipidomics was performed on brain cortex specimens from human subjects or mice or on apoE-containing lipoproteins from human CSF or mouse plasma using LC-MS as detailed in supplementary methods.
RESULTS
Sortilin interacts with apoE3 but not apoE4, to reduce brain Aβ42 levels
In mice, inactivation of Sort1, the gene encoding sortilin, results in brain accumulation of murine apoE due to impaired clearance of the protein by neurons lacking this apoE receptor [4]. Accumulation of murine apoE coincides with increased levels of Aβ40, suggesting a role for sortilin in neuronal catabolism of apoE-bound Aβ peptides [4]. While these findings implicated sortilin in apoE-dependent control of brain Aβ levels, they failed to address the relevance of this receptor for the increase in Aβ and the risk of AD seen in humans expressing apoE4 as compared with apoE3.
Now, we introduced the Sort1 defect (KO) into mice carrying a targeted replacement of the murine Apoe locus with genes encoding human apoE3 (E3; KO) or apoE4 (E4; KO) [5]. As shown for murine apoE before [4], loss of sortilin also caused accumulation of human apoE3 (Fig. 1A-B) and apoE4 (Fig. 1A, C) in cortex and hippocampus of KO mice compared with wild-types (WT). Sortilin deficiency did not impact transcript levels for apoE3 or apoE4 in the brain (Fig. 1D), nor the levels of apoE3 and apoE4 protein (Fig. S1A) or transcripts (Fig. S1B) in primary astrocytes from (E3;KO) or (E4;KO) mice as compared with WTs. Thus, increased brain levels of human apoE3 and apoE4 in KO mice were likely due to impaired neuronal clearance rather than increased astrocytic production of the apolipoproteins. These findings substantiated sortilin as major clearance receptor for murine and human apoE variants alike.
To test the impact of sortilin-mediated clearance of human apoE on brain Aβ levels, we crossed our mouse lines with the PDAPP strain expressing the human amyloid precursor protein APPV717F [6]. Brain accumulation of apoE3 or apoE4 in sortilin-deficient mice did not alter APP levels (Fig. 2A) but coincided with two-fold increased levels of Aβ40 in (E3;KO) and (E4;KO) mice compared with their respective controls (Fig. 2B). This observation supported a role for sortilin in apoE-dependent clearance of Aβ that also extends to human apoE variants. While no difference in accumulation of Aβ40 was evident comparing (E3;KO) and (E4;KO) mice (Fig. 2B), the situation was different for the more pathogenic variant Aβ42. Aβ42 levels were lower in brain cortex of WT mice expressing apoE3 as compared with apoE4, mirroring the higher prevalence of AD in humans with the APOE4 genotype (Fig. 2C). Loss of sortilin raised Aβ42 levels significantly in mice expressing apoE3 (E3;KO) but did not further increase the already high Aβ42 levels in (E4;KO) animals (Fig. 2C). The increase in Aβ42 in (E3;KO) animals was accompanied by a rise in sAPPβ levels (Fig. 2D), suggesting accelerated proteolysis of APP to sAPPβ and Aβ42 as the cause of increased accumulation of Aβ42 in apoE3 mice lacking sortilin. The same findings were obtained for Aβ40 (Fig. S2B) and Aβ42 (Fig. S2C) in the hippocampi of these mice.
Taken together, we documented a dual role for sortilin in control of amyloidogenic burden in the brain. It acts as a clearance receptor for Aβ peptides bound to apoE. This pathway regulates brain levels of Aβ40 but does not discriminate between apoE3 or apoE4. However, sortilin also interacts with apoE3 to selectively reduce the amyloidogenic processing of APP to Aβ42. This protective mechanism, that may decrease Aβ42 production by direct or indirect means, is lost with apoE4.
Interaction of sortilin and apoE3 controls brain levels of PUFAs and endocannabinoids
Next, we queried an interaction of sortilin with apoE in brain lipid homeostasis that may distinguish between apoE3 and apoE4 actions. Our strategy was based on the documented roles for apoE and the lipoprotein receptor sortilin in control of lipid metabolism [4, 7], and on the importance of neuronal lipid homeostasis for risk of AD (reviewed in [8]). To assure that alterations in lipids reflected primary mechanisms rather than secondary consequences of amyloidogenesis, this work was performed in mice lacking the PDAPP transgene.
Initially, we used LC-MS-based lipidomics to determine the levels of various lipid classes in brain cortices of (E3;KO) and (E4;KO) mice. When compared with WTs, we observed a distinct impact of genotypes on brain levels of total fatty acids (FA) and ω-3 polyunsaturated fatty acids (PUFAs), with levels being lower in (E3;KO) as compared with (E3;WT) mice (Fig. 3A). By contrast, levels of these lipid were always low in E4 compared with (E3;WT) animals, irrespective of the presence or absence of sortilin. This interaction of Sort1 and APOE in control of brain lipids was seen for ω-3 but not for ω-6 PUFAs (Fig. 3A), in line with the neuroprotective actions of ω-3 PUFAs in AD [9-11].
To delineate the relevance of sortilin and apoE3 for brain PUFA metabolism, we determined levels of bioactive PUFA derivatives in brains of our four mouse strains (Fig. 3B). We focused on endocannabinoids (eCBs), lipid-based neurotransmitters produced from PUFAs in neurons. Most eCBs act anti-inflammatory and neuroprotective, and alterations in eCB levels have been associated with AD [12]. Of the six eCBs and related lipids tested (Fig. 3B), levels of anandamide and linoleoyl-ethanolamide (LEA) were not impacted by genotypes. By contrast, levels of oleyl-ethanolamide (OEA) and palmitoyl-ethanolamide (PEA) were reduced in animals lacking sortilin, but irrespective of apoE genotype. Finally, levels of 2-arachidonoylglycerol (2-AG) and the eCB-like lipid synaptamide showed interaction between APOE and Sort1 as observed for brain Aβ42 levels before. In detail, physiological 2-AG levels were low in (E3;WT) mice but increased in (E3;KO) animals. By contrast, levels in E4 mice were always high compared with (E3;WT), irrespective of Sort1 genotype. The converse pattern was seen for synaptamide with physiological levels being high in (E3;WT) but lower in all other genotype combinations. The dependency of eCB levels on APOE genotype was substantiated in brain specimens of AD patients homozygous for APOE3 or APOE4 (Fig. 4). As in mice, levels of synaptamide were lower (p<0.05) while levels of 2-AG tended to be higher (p=0.057) in APOE4 as compared with APOE3 carriers. Levels of anandamide were also decreased in the APOE4 genotype (p<0.05), whereas OEA, LEA, and PEA levels were not impacted.
Sort1 and APOE3 safeguard an anti-inflammatory gene expression profile signaled by PPARs
Endocannabinoids exert their actions by signaling via G-protein coupled cannabinoid receptors CB1 and CB2 or by acting as ligands for peroxisome proliferator-activated receptors (PPARs). Because synaptamide does not engage CB1/2 [13], we explored the relevance of sortilin and apoE3 interaction in eCB metabolism by focusing on PPARs. Using a microarray-based strategy to assess transcript levels of 84 PPAR target genes, we identified 12 brain transcripts that showed dependence on Sort1 and APOE, being either higher or lower in (E3;WT) as compared with the other three genotypes (Fig. 5A). Affected transcripts included PPARγ (Pparg) and the retinoic X receptor (Rxrg), which forms heterodimers with PPARs; but also factors in lipid homeostasis, such as fatty acid transporter Slc27a4 and very long-chain acyl-CoA synthetase Slc27a5. Using qRT-PCR for PPAR targets Pparg, Mmp9, and Klf10, we substantiated an effect of sortilin on eCB-dependent gene transcription in E3 but not in E4 mice (Fig. 5B). PPARs are lipid sensors that suppress inflammatory responses by inducing an anti-inflammatory gene expression profile, an activity reducing neurodegeneration [14, 15]. In line with interaction of Sort1 and APOE3 in promoting PPAR activities, loss of sortilin in apoE3 mice or the presence of apoE4 (irrespective of Sort1 genotype) resulted in alterations in the mouse brains consistent with a pro-inflammatory state. These changes included decreased transcription of Vegf, but elevated transcript levels of Tnfα and Gfap in (E3;KO), (E4;WT), and (E4;KO) mice as compared with (E3;WT) (Fig. 5C). A potential pro-inflammatory state in the three mouse strains as compared with (E3;WT) was supported by increased immunosignals for GFAP in the brain (Fig. 5D-E).
Our data indicated interaction of Sort1 and APOE3 in safeguarding a neuroprotective metabolism and action of eCBs, potentially protecting the brain from inflammatory insults and decreasing amyloidogenic stimuli. This neuroprotective eCB action is lost in apoE3 mice that lack sortilin, raising brain Aβ42 levels. By contrast, this neuroprotective action of sortilin in brain lipid metabolism is not supported by apoE4, as PPAR activities are decreased and Aβ42 levels increased in E4 mice irrespective of Sort1 genotype.
Sortilin and apoE3 do not impact systemic metabolism of PUFAs
Anandamide and 2-AG are produced from arachidonic acid (ARA), while synaptamide is derived from docosahexaenoic acid (DHA). Mainly, ARA and DHA are supplied to the brain from the blood stream as FA or esterified to phospholipids. In the brain, ARA and DHA are re-esterified to membrane phospholipids or converted into bioactive metabolites, such as eCBs (reviewed in [16]). To query the impact of sortilin and apoE on PUFA transport into the brain, we quantified lipid levels in brain tissue, and in plasma and brain lipoproteins. Levels of DHA, but not of ARA, were reduced in the brains of (E3;KO), (E4;WT), and (E4;KO) compared with (E3;WT) mice (Fig. S3A). A similar trend in reduction of DHA (p=0.1) but not ARA levels was seen in brain tissues from AD patients with APOE4/4 as compared with APOE3/3 (Fig. S3B). Altered brain levels of DHA were not due to differential association of lipids with lipoproteins containing apoE3 or apoE4 as plasma lipoprotein profiles were indistinguishable comparing all four mouse genotypes (Fig. S4A), as were the levels of DHA and ARA in apoE-containing lipoproteins isolated from plasma of these mice (Fig. S4B). Also, levels of DHA and ARA were comparable in apoE3- and apoE4-containing brain lipoproteins isolated from human cerebrospinal fluid (Fig. S4C-F). Finally, transcript levels of enzymes in neuronal metabolism of 2-AG and synaptamide were not affected by Sort1 or APOE genotype as shown by qPCR on mouse brain extracts (Fig. S5).
ApoE4 disrupts cell surface recycling of sortilin
So far, our studies failed to identify changes in biosynthesis or extracellular transport that may explain why alterations in DHA and eCB metabolism in E4 mice resembled defects in apoE3 animals lacking sortilin. Thus, we focused on the hypothesis that apoE4 may disrupt the activity of sortilin, rendering E4 brains essentially sortilin-depleted. This hypothesis was based on the propensity of apoE4 to impair intracellular sorting and thereby activity of several cell surface receptors [17-19].
To query whether trafficking of sortilin was differentially impacted by apoE variants, we tested its co-localization with apoE3 or apoE4 in CHO cells using proximity ligation assays (PLA). Co-localization of sortilin with apoE3 was seen in a scattered vesicular pattern throughout the cytoplasm, whereas co-localization with apoE4 showed a distinct pattern close the cell membrane (Fig. 6A). To substantiate this differential impact of apoE4 on sortilin sorting, we established the trafficking path of the receptor in CHO cells by labeling sortilin molecules on the cell surface with antibodies and following their subsequent intracellular route using immunocytochemistry (Fig. 6B). Within 15 minutes, labeled receptors internalized from the cell surface (Fig. 6B, panel surface labeling) into intracellular compartments (Fig. 6B, panel internalization). To interrogate recycling of these internalized receptors, we subsequently treated the cells with dynasore, an inhibitor of endocytosis to block continuous re-entry of recycled receptors into the cells. Application of dynasore resulted in the accumulation of labeled sortilin molecules at the plasma membrane, confirming cell surface recycling of internalized (antibody-labeled) receptors (Fig. 6B, panel recycling). This recycling path of sortilin was not impacted by the presence of apoE3 in the culture medium (Fig. 6C). However, in the presence of apoE4, internalized receptors failed to re-appear on the cell surface and remained largely intracellular (Fig. 6C). To more accurately quantify the extent of sortilin recycling in the presence of apoE3 versus apoE4, we biotinylated sortilin molecules on the surface of CHO cells and followed their internalization and recycling fate as schematized in Fig. 6D. Recycled receptors accumulating at the cell surface were stripped of their biotin tag by glutathione treatment to only retain a biotin label in the intracellular (non-recycling) receptor pool. The extent of receptor recycling was determined by subtracting the amount of biotinylated receptors purified on streptavidin beads after 60 minutes of recycling from the total amount of biotinylated receptors internalized at the beginning of the experiment. In these studies, the amount of receptors recycling back to the cell surface was reduced by 50% in apoE4-treated as compared to apoE3-treated cells (Fig. 6E and F).
DISCUSSION
We propose a novel concept for the neuroprotective metabolism of PUFA in the brain. It involves interaction of sortilin with apoE3 to support neuronal uptake and action of DHA and eCBs, providing anti-inflammatory gene expression und suppressing noxious insults that increase Aβ42 production. Loss of sortilin in apoE3 mice compromises this neuroprotective PUFA uptake pathway, increasing pro-inflammatory gene expression and amyloidogenic processing. Similar detrimental effects are seen in wild-type mice in the presence of apoE4 as it disrupts receptor trafficking.
In mice, loss of sortilin reduces neuronal uptake of apoE and causes alterations in brain lipid homeostasis also seen in apoE KO animals (i.e., accumulation of sulfatides) [4]. These data suggested sortilin as a key player in apoE-dependent lipid metabolism in the brain. Our new results substantiate this role by identifying the interaction of sortilin with apoE3 in control of neuronal PUFA metabolism and signaling. Likely, this neuroprotective lipid pathway explains the anti-amyloidogenic interaction of apoE3 and sortilin (Figs. 2 and S2) as the same interaction between Sort1 and APOE genotypes, as for Aβ42, is also seen for brain levels of DHA and several eCBs (Fig. 3). Importantly, the same effect of APOE4 on DHA and eCBs levels is seen in AD patients, substantiating the clinical relevance of our findings from mouse models (Fig. 4).
DHA and its bioactive metabolites are regulators of brain health, linking inflammation with neurodegeneration. DHA is the most abundant brain ω-3 PUFA. It regulates numerous cellular processes, especially the resolution of brain inflammation [16, 20]. Low levels of DHA in plasma or brain correlate with human AD [21-23] and dietary DHA intake lowers Aβ levels in mice [9-11]. A similar role in control of inflammation in the ageing brain has been suggested for anandamide and 2-AG. Anandamide levels decrease in the brain of patients [24, 25] and mice with AD [26]. Conversely, levels of 2-AG increase in AD patients [27] and mouse models [28], and increased 2-AG signaling exacerbates synaptic failure [29]. Although little is known about synaptamide, it is considered a bioactive mediator of DHA as it recapitulates the protective actions seen with DHA [30]. We document a pathological increase in brain levels of 2-AG and a concomitant decrease in DHA, anandamide, and synaptamide in humans and mice with apoE4, and in apoE3 mice lacking sortilin, identifying the relevance of Sort1 and APOE genotypes for brain PUFA homeostasis.
We focused on the relevance of sortilin and apoE3 interaction for DHA and eCB signaling by investigating PPARs. Transcript levels of PPAR-γ increase in AD patients to counteract neuroinflammatory insults [31]. Non-steroidal anti-inflammatory drugs ameliorate AD-related processes due to their ability to stimulate PPAR and inhibit inflammatory responses [32, 33]. In line with a function for sortilin and apoE3 in PPAR actions, we observed a concordant dysregulation of transcriptional targets in E4 mice and in E3 animals lacking sortilin (Fig. 5A-B). The relevance of Sort1 and APOE3 to counteract brain inflammation was supported by an increase in pro-inflammatory markers GFAP and TNFα (Fig. 5C-E). GFAP indicates brain inflammation in rodent AD models [34], and increases in GFAP levels in AD rats are reverted by eCB application [35]. Increased levels of TNFα are also linked to AD pathology [36, 37]. Another sign of a potentially pro-inflammatory state in E4 mice and in (E3;KO) animals is down-regulation of VEGF (Fig. 5C). VEGF suppresses inflammatory brain response [38] and reduces neurodegeneration in mouse models of AD [39, 40].
Sortilin facilitates cellular uptake of apoE3 and apoE4 equally well, both in vitro [4] and in vivo (Fig. 1A-C). Also, levels of PUFA bound to apoE3- or apoE4-containing lipoprotein in plasma or brain are comparable (Fig. S4). Thus, apoE variants likely do not discriminate the types of PUFA taken up into neurons via sortilin. Rather it is the detrimental effect of apoE4 on trafficking of sortilin that impairs this neuronal uptake route for lipids (Fig. 6). Following endocytic uptake, apoE3 recycles to the cell surface for re-secretion. This route is not taken by apoE4 that accumulates in the endocytic pathway [17, 18, 41]. Impaired recycling of apoE4 disrupts cell surface re-exposure of apoE binding proteins, as shown for APOER2 [19] or the insulin receptor [42], and reversal of impaired recycling has been suggested as therapeutic approach for AD risk imposed by apoE4 [43]. We show a similar deleterious impact of apoE4 on sortilin trafficking (Fig. 6), likely explaining the loss of neuroprotective actions of this receptor in eCB metabolism and Aβ42 accumulation in the apoE4 brain.
In conclusion, we identify a novel pathway linking neuronal apoE handling by sortilin with the well-known neuroprotective actions of PUFA in the brain, and we provide a possible molecular explanation for the risk of AD seen with apoE4.
Statement of conflict
The authors have no competing interest to declare.
Acknowledgement
We are indebted to T. Pasternack, K. Kampf, C. Kruse, M. Kahlow, and S. Ehret for expert technical assistance. Studies were funded in part by the ERC (BeyOND No. 335692), the Helmholtz Association (AMPro), the Alzheimer Forschung Initiative (#18003), and the Novo Nordisk Foundation (NNF18OC0033928) to TEW.