Co-aggregation with Apolipoprotein E modulates the function of Amyloid-β in Alzheimer's disease

doi:10.21203/rs.3.rs-2208663/v1

Download PDF

Article

Co-aggregation with Apolipoprotein E modulates the function of Amyloid-β in Alzheimer's disease

https://doi.org/10.21203/rs.3.rs-2208663/v1

This work is licensed under a CC BY 4.0 License

Journal Publication

published 31 May, 2024

Read the published version in Nature Communications →

Version 1

posted

You are reading this latest preprint version

Which isoforms of apolipoprotein E (apoE) we inherit determine our risk of developing late-onset Alzheimer’s Disease (AD), but the mechanism underlying this link is poorly understood. In particular, the relevance of direct interactions between apoE and amyloid-β (Aβ) remains controversial. Here, single-molecule imaging shows that all isoforms of apoE associate with Aβ in the early stages of aggregation and then fall away as fibrillation happens. ApoE-Aβ co-aggregates account for ~ 50% of the mass of soluble Aβ aggregates detected in the frontal cortices of homozygotes with the higher-risk APOE4 gene. Our results connect inherited APOE genotype with the risk of developing AD by demonstrating how, in an isoform- and lipidation-specific way, apoE modulates the aggregation, clearance and toxicity of Aβ. Selectively removing non-lipidated apoE4-Aβ co-aggregates enhances clearance of toxic Aβ by glial cells, and reduces inflammation and membrane damage, demonstrating a clear path to AD therapeutics.

Biological sciences/Neuroscience/Molecular neuroscience

Biological sciences/Biophysics/Single-molecule biophysics

Biological sciences/Neuroscience/Diseases of the nervous system/Alzheimer's disease

Inherited variation in the sequence of apolipoprotein E (apoE) is the greatest genetic risk factor for late-onset Alzheimer’s Disease (AD), which accounts for ~ 95% of all AD cases¹. The APOE gene has three alleles: APOE2, APOE3, and APOE4. Compared to the most common APOE3 form, APOE2 is neuroprotective², while APOE4 increases AD risk; APOE4 homozygotes are 15-fold more susceptible to AD than APOE3 homozygotes^3,4, and disease begins several years earlier in APOE4 carriers than those with APOE3 or APOE2³.

Attempts to establish how apoE influences AD risk have focused on its effect on the pathological amyloid-beta (Aβ) peptide^1,5. Deposits of Aβ aggregates in the central nervous system (CNS) are a hallmark of AD⁶, and disturbed Aβ homeostasis is one of the earliest events in AD pathogenesis^7,8. This dysfunction induces synaptic and axonal damage as well as tau seeding and spreading, leading to neurodegeneration in AD^9–11. There is compelling evidence that apoE4 enhances Aβ pathology⁵: carriers of the APOE4 gene have more Aβ deposits in their CNS than non-carriers^1,12, exhibit amyloid positivity earlier in life¹³, and experience a faster-growing Aβ burden¹⁴. The isoforms of the apoE protein seem to differentially affect the levels of Aβ in the CNS, which may explain their differential AD risk profiles. Lipidation of apoE may also play a role in AD. Although mostly lipidated in the human CNS^5,15, poorly or non-lipidated apoE increases Aβ pathology¹⁶, and the risk of developing AD¹⁷. Observational studies show that apoE influences the relationship between Aβ and cognitive decline in AD^18–20. However, it is unclear how the different isoforms and lipidation states of apoE affect Aβ aggregation, clearance, and aggregate-induced neurotoxicity at the molecular level.

The faster Aβ plaque deposition in carriers of APOE4^14,21 has led to several hypotheses for apoE4’s molecular role in AD. One possibility is that apoE4 promotes more aggregation of Aβ than the other isoforms by interacting directly with Aβ when the two meet in the extracellular space. The discovery of co-deposited apoE in AD amyloid plaques provided early circumstantial evidence for this idea^15,22,23, but biophysical studies are equivocal: apoE can either speed up or slow down Aβ aggregation in vitro, depending on the conditions²⁴. Targeting non-lipidated apoE in amyloid plaques with an antibody can significantly reduce Aβ pathology in transgenic mouse models^15,22 as well as reduce Aβ-induced tau seeding and spreading²⁵. However, studies disagree on whether soluble forms of apoE and Aβ associate significantly in brain tissue^17,26. The role of apoE-Aβ interactions in Aβ clearance is also unclear: while apoE might traffic Aβ out of the interstitial brain fluid^21,27, other work suggests that apoE and Aβ instead compete for clearance-mediating receptors²⁶.

In this study, we used single-molecule imaging to monitor how apoE affects the form, function, and clearance of Aβ aggregates at different stages of maturity. We show that apoE influences the structure and composition of Aβ aggregates during oligomerization, and that apoE-Aβ interactions tune disease-relevant functions of Aβ in an isoform- and lipidation-specific manner. Our work identifies a role for transient, non-lipidated apoE-Aβ co-aggregates in modulating Aβ deposition and neurotoxicity, and thereby connects APOE genotype inheritance with the risk of developing sporadic AD.

Soluble apoE-Aβ Co-aggregates Form Isoform-independently, but Accumulate Isoform-dependently in AD Brains

We began our study by investigating how apoE interacts with Aβ along its aggregation pathway to fibrils. Aggregating Aβ comprises a dynamic, heterogeneous mixture of species with different sizes, shapes, and properties, and small sub-populations may disproportionately contribute to AD^28,29. We reasoned that apoE could influence AD risk by interacting with Aβ aggregates that are transient and/or rare, which might explain why previous attempts to study association by taking snapshots of the bulk mixture have painted an inconsistent picture. We therefore aggregated Aβ42 in vitro at a concentration (4 µM) that would give rise to fibrils, in the presence and absence of near-physiological concentrations of each non-lipidated and lipidated apoE isoform (~ 80 nM)^30,31.

Assaying fibril formation using ThT fluorescence confirmed that all reactions produced fibrils and that the presence of non-lipidated apoE slowed aggregation, while the presence of lipidated apoE sped the reaction up (Fig. 1A, Figure S1); lipid particles are known to accelerate the aggregation of amyloidogenic proteins by providing a surface for primary nucleation^32–34. In order to sample the heterogeneity within each aggregation pathway, we characterized individual aggregates at different stages of the reaction. We imaged each aggregation reaction at the end of the lag phase (t₁), the middle of the growth phase (t₂), and the plateau phase (t₃), using single-molecule pull-down (SiMPull)³⁵ (Fig. 1B). In this assay (Fig. 1C), Aβ42 is captured using a surface-tethered 6E10 antibody and imaged using two-color total internal reflection fluorescence (TIRF) after adding primary detector antibodies for Aβ (Alexa-Fluor-647-labeled 6E10) and apoE (Alexa-Fluor-488-labeled EPR19392) (Figure S2). Using the same monoclonal antibody to sandwich Aβ aggregates renders unreacted monomers undetectable because they only contain one epitope.

Characterizing individual aggregates rather than their ensemble average allowed us to extract properties of the heterogeneous population including size, shape, and composition; aggregates containing apoE and Aβ42 should be colocalized in both detection channels. These images (Fig. 1D-F) revealed that all types of apoE coaggregate with Aβ42 in the early stages (t₁ and t₂) of aggregation, but that co-aggregates disappear as the reaction reaches completion. There was no significant isoform dependence in the extent of colocalization, whether based on the aggregate number (Fig. 1G), or intensity (Fig. 1H). These findings are independent of the antibody used to detect apoE (Figure S3), and there is no colocalization when isotype-control detection antibodies are used (Figures S4 and S5), suggesting minimal contributions from non-specific binding. Using the intensity as a proxy for the amount of protein present in each aggregate suggests that ~ 75–100% of aggregate mass at the end of the lag phase is in co-aggregates, which falls to ~ 30%-60% in the growth phase, and ~ 0% by the plateau phase. Although non-lipidated apoE colocalizes with Aβ42 at slightly higher levels, the main effect of lipidation was on the apparent size of co-aggregates (Fig. 1I): those containing non-lipidated apoE (mean diameter ~ 500–900 nm) were much larger than those containing lipidated apoE (mean diameter ~ 200–250 nm or less; in diffraction-limited imaging, any aggregates smaller than the diffraction limit will appear in this range).

Our finding that non-lipidated apoE forms large soluble co-aggregates with Aβ supports the idea that apoE in this form can stabilize soluble species formed early on in aggregation and thus inhibit Aβ fibrillation³⁶. Because we used apoE in sub-stoichiometric amounts (Aβ: apoE 50:1), it is unlikely to slow down aggregation by sequestering Aβ monomers. ApoE more likely interacts with Aβ42 in its earliest stages of aggregation; apoE did not associate with pre-formed Aβ42 aggregates (Figure S6). The high fluorescence intensities of these soluble co-aggregates indicate high effective protein concentrations, which may seem incompatible with decreased fibrillization rates. However, it has recently been shown that locally concentrating Aβ42 in condensates can significantly slow its aggregation³⁷. The fact that fibrils have shed all associated apoE might indicate that elongating heteronuclei is less energetically favorable than elongating homonuclei. Importantly, the lack of any isoform dependence means that co-aggregation alone cannot explain the APOE dependence of AD.

To determine whether these apoE-Aβ co-aggregates are disease-relevant, we isolated Aβ aggregates from the postmortem frontal cortices of six AD patients (Supplementary Table 1). We chose a method that gently extracts soluble aggregates by soaking³⁸, in preference to tissue homogenization, which extracts insoluble, predominantly fibrillar aggregates that are mostly inert³⁹. We imaged the extracts from three homozygous APOE4 and three homozygous APOE3 AD patients using SiMPull (Fig. 2A). All samples yielded Aβ aggregates both with and without associated apoE; extracts from APOE4 carriers contained more of both types of aggregates than extracts from APOE3 carriers (Fig. 2B). This finding supports the previous report that APOE4 carriers have more soluble Aβ aggregates than APOE3 carriers^40,41. Unlike our in vitro aggregations - where the initiation of aggregation is synchronized for all monomers - only ~ 5% of Aβ aggregates in APOE4 homozygotes and ~ 1% of aggregates in APOE3 homozygotes are co-aggregates (Fig. 2C). This difference probably results from the transience of co-aggregates and the fact that Aβ and apoE are replenished and cleared over years in the CNS. Extracted aggregates therefore reflect a broad sample of the reaction pathway, e.g. recently formed soluble aggregates alongside mature fibrils. Although few in number, some of the ex vivo co-aggregates are large (Fig. 2E) and contribute disproportionately to the fluorescence intensity. Using the intensity as a proxy for total amounts of protein suggests that co-aggregates comprise 40–60% and 10–35% of the Aβ-aggregate mass in APOE4 and APOE3 carriers, respectively (Fig. 2D). These data, therefore, reveal a large isoform dependence in the accumulation of co-aggregates in AD brains.

Glial Uptake of Soluble apoE-Aβ Co-aggregates is Isoform- and Lipidation-dependent

Given that all isoforms of apoE, whether lipidated or not, had co-aggregated to similar extents with Aβ42, we next asked whether differential clearance might explain the isoform-dependent accumulation of apoE-Aβ co-aggregates in AD brains. We quantified the uptake of aggregates by two types of cells: i) immortalized murine microglial BV-2 cells, and ii) astrocytes isolated from the cerebral cortex of humans (Fig. 3A-C). BV-2 cells and human astrocytes take up soluble and insoluble Aβ aggregates^42,43, and dysfunction of microglia and astrocytes is implicated in AD^44,45. Alongside monitoring aggregate clearance by these cells, we also measured the resulting neuroinflammation. Protein aggregates cause neuroinflammation by disrupting CNS homeostasis⁴⁶, and neurotoxic Aβ aggregates are known to promote pro-inflammatory pathways in microglia⁴⁴ and astrocytes⁴⁵.

To isolate the role of co-aggregates, we compared the uptake of soluble aggregates from t₁ (≥ 75% co-aggregates) and fibrils from t₃ (~ 0% co-aggregates) by microglia and astrocytes over a 24-hour period. Both cell types took up more soluble co-aggregates than fibrils prepared in the presence of lipidated apoE, but not non-lipidated apoE (Fig. 3D-E). Uptake was also isoform-specific for soluble non-lipidated co-aggregates, which were internalized ~ 2-3-fold more efficiently if they contained apoE2 rather than apoE4 (Figure S7). Co-aggregate clearance is also highly dependent on lipidation status: lipidated co-aggregates were internalized to a much greater extent than non-lipidated co-aggregates (Figure S8).

To assess how much neuroinflammation results from clearing co-aggregates, we assayed the secretion of the pro-inflammatory cytokine, TNF-α, using an enzyme-linked immunosorbent assay (ELISA) (Fig. 4A). Soluble aggregates were significantly more inflammatory than fibrils if apoE was absent from the aggregation reaction or present in a non-lipidated form. However, the presence of lipidated apoE reduced the inflammation caused by soluble co-aggregates to levels similar to fibrils (Fig. 4B-C). As with clearance, neuroinflammation was significantly isoform-dependent only for non-lipidated co-aggregates, which induced significantly more TNF-α secretion if they contained apoE4 rather than apoE2 (Figure S9). The lipidation status of apoE also affected neuroinflammation: co-aggregates induced higher levels of TNF-α release when they were formed in the presence of non-lipidated forms of apoE (Figure S10).

These responses of microglia and astrocytes to co-aggregates begin to tease out isoform- and lipidation-dependent patterns that hint at a co-aggregate-mediated link between APOE genotype and AD risk. In the presence of non-lipidated apoE, soluble co-aggregates were highly inflammatory, but lipidation enhanced their uptake and suppressed inflammation. This pattern suggests a role for soluble, non-lipidated apoE-Aβ co-aggregates. Within this group, the expected isoform dependence was apparent, i.e. non-lipidated apoE4-Aβ co-aggregates were the least well cleared and most inflammatory of all Aβ species.

Soluble, Non-lipidated apoE4-Aβ Co-aggregates are Toxic and Impair Clearance by Glial Cells

It is widely reported that aggregated Aβ is toxic, and that its toxicity depends on the size, structure, and composition of the aggregate. To check whether apoE modulates the toxicity of Aβ aggregates in an isoform or lipidation-specific way, we measured the toxicity of Aβ at different stages of aggregation. We assayed this property in two ways: i) the ability to permeabilize lipid membranes and cause Ca²⁺ influx⁴⁷ and ii) toxicity to the SH-SY5Y human cell line, leading to the release of lactate dehydrogenase (LDH) (Fig. 5A and C). We again compared soluble co-aggregates from t₁ to fibrils from t₃ - which have shed apoE - in order to isolate the role of co-aggregates. Soluble co-aggregates induced higher Ca²⁺ influx and LDH release than fibrils when prepared in the presence of non-lipidated apoE, but not lipidated apoE (Fig. 5B and 5D). Echoing our results on uptake and neuroinflammation, the toxicity of co-aggregates was isoform-dependent (apoE4-Aβ being more toxic than apoE2-Aβ) for Aβ aggregates containing non-lipidated apoE, but not lipidated apoE (Figure S11A and S12A). Non-lipidated apoE4-Aβ co-aggregates permeabilized lipid membranes more potently (Figure S11B) and induced more LDH secretion (Figure S12B) than the corresponding lipidated forms. This lipidation dependence was absent for soluble apoE2-Aβ and apoE3-Aβ co-aggregates and fibrils prepared in the presence of all apoE isoforms.

These toxicity assays painted a similar picture: soluble co-aggregates were better at permeabilizing lipid bilayers and more toxic to neuroblastoma cells than fibrils, with both lipidation- and isoform-dependent features. Lipidated co-aggregates were less toxic than non-lipidated co-aggregates, and among the latter, toxicity was isoform dependent. Non-lipidated apoE-4Aβ co-aggregates stand out as the most toxic and damaging to lipid membranes.

Having established that non-lipidated apoE-Aβ co-aggregates were especially inflammatory and toxic, we wondered whether their selective removal might rescue Aβ-induced dysfunction. To test this hypothesis, we immunoprecipitated non-lipidated apoE4-Aβ co-aggregates from a 1:1 mixture of lipidated and non-lipidated co-aggregates using HAE-4, an antibody specific for non-lipidated apoE4^15,22,25, and an isotype control (Fig. 6A-B). Cellular uptake and neuroinflammation assays on the mixture and supernatant showed that removing non-lipidated co-aggregates does indeed enhance Aβ clearance and reduces Aβ-associated inflammation in both microglia (Fig. 6C-D) and astrocytes (Fig. 6E-F). Immunoprecipitating non-lipidated apoE4-Aβ co-aggregates also reduced membrane damage (Fig. 6G) and LDH release (Fig. 6H). It is counterintuitive that removing the non-lipidated half of the aggregates increases the total uptake of Aβ (by ~ 50% for astrocytes and ~ 100% for microglia), even though there is ~ 50% less total Aβ. However, exclusively lipidated apoE4-Aβ co-aggregates (Fig. 3D-E) are taken up at well over double the levels of a 1:1 mixture of lipidated and non-lipidated co-aggregates. This implies that the inflammation caused by contact with non-lipidated co-aggregates impairs the ability of astrocytes and microglia to clear lipidated aggregates. Sparing cells from this insult by removing non-lipidated co-aggregates therefore increases the clearance of lipidated co-aggregates.

Soluble apoE4-Aβ Co-aggregates Detected in AD Brains are Poorly Lipidated

Finally, to corroborate the putative role of non-lipidated apoE co-aggregates, we asked whether Aβ-aggregates containing non-lipidated apoE are over-represented in postmortem AD brain tissue. We were unable to source a specific antibody to either lipidated or non-lipidated apoE3 with sufficiently low non-specific or cross binding, so these experiments were conducted solely on samples from APOE4/4 homozygotes. Immunoprecipitating non-lipidated apoE from these brain extracts using the HAE-4 antibody (Fig. 7A) disproportionately removed the largest co-aggregates (Fig. 7B). The drop in colocalization after immunoprecipitation (Fig. 8C-D) implies that ~ 80% of co-aggregates trapped in the neural tissue of AD patients contain poorly lipidated apoE; this is striking given that only a minority of apoE in the CNS is non-lipidated.

It is widely thought that soluble Aβ aggregates are key molecular players in AD: they trigger neuronal dysfunction, impair synaptic plasticity and their abundance correlates with severity of neurodegeneration in AD^48,49. Attempts to reconcile this hypothesis with the link between APOE inheritance and late-onset AD risk have so far been unsuccessful. This study identifies a potential role for direct interactions between apoE and soluble Aβ aggregates in forging that link. Starting from the premise that apoE could affect AD risk by interacting with Aβ aggregates that are either rare or transient, we sampled a broad cross-section of the aggregation equilibrium. We found that apoE-Aβ co-aggregates are transient, but not rare: the journey to apoE-free fibrils passed through highly populated apoE-containing intermediates in vitro, irrespective of isoform or lipidation of apoE. In contrast, there were stark isoform- and lipidation-dependent differences in the abundances of ex vivo co-aggregates. Co-aggregates comprise more of the soluble aggregated Aβ detected in the frontal cortices of APOE4/4 patients (~ 50%) than APOE3/3 patients (~ 20%). About 80% of ex vivo co-aggregates detected from APOE4 homozygotes contain the less-common, non-lipidated form of apoE.

Rather than deriving from aggregation kinetics or thermodynamics, the increased deposition of soluble, non-lipidated co-aggregates in APOE4/4 brains seems to have a cellular basis. In the CNS, most Aβ aggregates are removed for degradation by glial cells via phagocytosis, pinocytosis, or receptor-mediated endocytosis^42,50. Exposure to aggregates stresses microglia and astrocytes, which then release pro-inflammatory cytokines such as TNF-α, damaging nearby neurons and promoting neurodegeneration^51,52. We found that co-aggregation with lipidated apoE enhanced Aβ uptake, while non-lipidated apoE4-Aβ co-aggregates were the slowest cleared and most inflammatory of all soluble aggregates. Our results suggest that apoE interacts with newly formed soluble aggregates in the extracellular space; if the co-aggregate contains lipidated apoE, this accelerates its clearance and avoids Aβ-induced toxicity. On the other hand, co-aggregates containing non-lipidated apoE4 are poorly cleared and inflame glial cells, which impairs the clearance of other soluble Aβ aggregates. This dynamic would lead to greater deposition of Aβ over time in individuals with APOE4 genotypes. The fact that apoE4 is less lipidated in vivo^53,54 than apoE2 or apoE3 may exacerbate this vicious circle.

This apoE-mediated mechanism for Aβ deposition, combined with co-aggregate toxicity may influence disease progression in AD. Non-lipidated apoE-Aβ co-aggregates are more likely to be deposited than lipidated co-aggregates, but are also more toxic. This may explain why deleting the ATP-binding cassette transporter 1 (ABCA1) - which regulates apoE lipidation in the CNS¹⁶ – increases the deposition of Aβ plaques^55,56, while overexpressing or upregulating ABCA1 relieves cognitive impairment caused by Aβ and reverses memory deficits in transgenic animals⁵⁷. The effect of lipidation, alongside the isoform-dependence of co-aggregate toxicity connect the APOE4 genotype with increased amyloid load, as well as faster AD onset and progression, via a possible culprit: soluble, non-lipidated apoE4-Aβ co-aggregates. Importantly, mature Aβ fibrils do not contain apoE and are functionally indistinguishable, whichever form of apoE was present during aggregation. This finding strengthens our hypothesis that apoE contributes to AD risk by modulating the behaviour of soluble Aβ aggregates.

Our work suggests further avenues for investigating and treating late-onset AD. We show that removing non-lipidated apoE4 co-aggregates reduces inflammation, enhances Aβ uptake by glial cells, and protects from Aβ-induced membrane disruption and subsequent Ca²⁺ dysregulation. This may form the molecular basis of immunotherapy with the HAE-4 monoclonal antibody raised against non-lipidated human apoE^15,22. We found opposing trends in the isoform-dependence of internalization and membrane permeabilization of co-aggregates, suggesting that uptake may be mainly receptor-mediated. Because the receptor binding capacity of apoE depends on its lipidation^58,59, the fact that lipidation of co-aggregates enhances their uptake supports this idea. ApoE’s influence on endocytosis is well studied, with roles for other AD risk factors such as LDL⁶⁰, TREM2⁶¹ and PICALM⁶², while LRP1 and VLDLR are putative receptors for the transport of apoE-Aβ complexes²⁷. Validating specific receptor(s) which play a significant role in the internalization of large soluble co-aggregates could present opportunities to precisely target apoE in AD. Modulating the formation or uptake of non-lipidated, soluble co-aggregates in an isoform-specific way could eliminate the AD-specific role of apoE without impeding its essential functions in lipid transport. Enhancing apoE lipidation could also provide a new intervention strategy for late-onset Alzheimer’s disease, specifically for APOE4 carriers, because apoE4 is less lipidated than apoE2 or apoE3^5,54.

Aggregation of recombinant Aβ42

Solutions of monomeric recombinant Aβ42 peptide (Stratech, Cat. No. A-1167-2) were diluted in phosphate-buffered saline (PBS) buffer to a concentration of 4 µM and aggregated in a 96-well half-area plate (Corning, Cat. No. 3881) at 37°C, under quiescent conditions in the absence and presence of non-lipidated or lipidated apolipoprotein E (apoE) isoforms E2, E3 or E4 (Peprotech, Cat. No. 350 − 12, 350-02 or 350-04) at concentrations of 80 nM (lipidated apoE2, apoE3 and apoE4, and non-lipidated apoE2 and apoE3) or 20, 40 and 80 nM (non-lipidated apoE4). The aggregation process was monitored using 20 µM Thioflavin T dye (Sigma, Cat. No. T3516) using a plate reader (Clariostar Plus, BMG Biotech).

To analyze the species formed during Aβ42 aggregation and co-aggregation with apoE, samples of the aggregation mixture (aggregated without ThT) were taken at the end of the lag phase (t₁), ~ 15 minutes' incubation (30 min and 10 min for co-aggregation with non-lipidated and lipidated apoE respectively), middle of the growth phase (t₂) ~ 30 minutes' incubation (45 min and 20 min for co-aggregation with non-lipidated and lipidated apoE respectively) and at the plateau phase (t₃) ~ 90 minutes' incubation (120 min for co-aggregation with both non-lipidated and lipidated apoE).

Lipidation of recombinant apoE: ApoE was lipidated using a previously published protocol²⁶. All the isoforms of apoE were lipidated using a cholate dialysis method using 1-palmitoyl-2-oleoyl-glycerol-3-phosphocholine 16:0–18:1 PC (POPC) (Avanti Lipids, 850457C) and Cholesterol (Avanti Lipids, 700100P) in a molar ratio 1:90:5 (apoE: POCP: cholesterol). These ratios of apoE and lipids were selected to mimic the physiological lipid composition of HDL-like apoE particles. Lipids can be directly incorporated into recombinant apoE, but the sodium cholate dialysis method has been shown to produce lipidated apoE homogeneously in a reproducible manner.

First, recombinant apoE was suspended in 600 µL of apoE buffer [20 mM phosphate buffer (Sigma) containing 50 mM NaCl (Sigma, Cat. No. S7653), 1 mM dithiothreitol (DTT, Sigma, Cat. No. 10708984001), and 1 mM ethylenediaminetetraacetic acid (EDTA, Sigma, Cat. No. E9884) at pH 7.4]. In parallel, stock solutions of 10 mg/mL 16:0–18:1 PC (Avanti Lipids, Cat. No. 850457C), and 10 mg/mL cholesterol (Avanti Lipids, Cat. No. 700100P) in chloroform were made. Then 16:0–18:1 PC (18.3 µL) and cholesterol (2 µL) were mixed in a glass vial at a molar ratio of 90:5 and the chloroform was then removed under vacuum in a desiccator overnight. Then 40 µL of Dulbecco's phosphate-buffered saline buffer (DPBS, Thermo Fisher, Cat. No. 141901440) with 0.5 mM DTT (Sigma) was added to each vial to resuspend the lipid at a total lipid concentration of 5 µg/µL, which was then vortexed three times for 10 min at 2 minute intervals. Then the solution left to stand for 30 minutes at room temperature. Then, sodium cholate (50 mg·mL^− 1, Sigma) was slowly added to the lipid solutions with vortexing until the solution was free of turbidity. The apoE solution was then added and the mixture vortexed three times for 10 minutes, at 5 minute intervals and left to stand for 1 hour. Then the entire solution was dialyzed using Slide-A-Lyzer Dialysis Cassettes with a 10 kDa -cut off (Thermo Fisher Cat 66380) with apoE buffer for 72 h, (with a buffer change every 24 h) at 4°C. After this, the apoE solutions were purified by gel filtration chromatography (Superose 6 Increase 3.2/300) at 4°C, and the fractions containing lipidated apoE were combined and concentrated using an Amicon Ultra 15 mL 10 kDa cut-off centrifugal filter unit (Merck Millipore, Cat. No. UFC901024) at 3000 g for 20 min at 4°C. The concentration of lipidated apoE was determined by absorbance measurements at 280 nm using an extinction coefficient of 44,460 M^− 1·cm^− 1. Finally, the samples of lipidated apoE were stored at 4°C. Lipidation of all the apolipoproteins were performed in parallel using the same lipid–cholesterol suspension.

Postmortem brain tissue

Flash-frozen brain tissues from six Alzheimer's disease patients (three homozygous APOE4 and three homozygous APOE3 AD patients) were obtained from the Newcastle brain tissue resource (NTBR) under a Material Transfer Agreement. Tissue was processed at the NTBR, where regions of interest were removed from the frontal cortex (Brodmann area 9) and immediately frozen at -80°C. Data relating to supplied tissue were released on an anonymized basis. Ethical approval was granted through the Newcastle and North Tyneside Research Ethics Committee.

Supplementary Table 1. Properties of postmortem brain tissue samples.

Sex	Age	Postmortem time before freezing (hours)	Braak stage	APOE genotype
F	70	24	6	4/4
M	90	69	6	4/4
F	80	10	6	4/4
F	58	70	6	3/3
F	84	26	6	3/3
F	88	57	6	3/3

Extraction of soluble protein aggregates from postmortem tissue

Soluble Aβ-containing aggregates were extracted from human tissue following a recently published protocol with minor changes³⁸. In summary, the supplied tissues were chopped into 300 mg pieces using a razor blade and incubated with 1 mL of artificial cerebrospinal fluid (aCSF) buffer (120 mM NaCl, 2.5 mM KCl, 1.5 mM NaH₂PO_4, 26 mM NaHCO₃, 1.3 mM MgCl₂, pH 7.4) at 4°C for 30 min under gentle agitation. The mixture was then centrifuged for 10 min at 2,000 g at 4°C. The upper ~ 80% of the supernatant was further centrifuged for 110 min at 14,000 g at 4°C. The upper ~ 80% of the supernatant was collected and dialyzed using a 2 kDa molecular-weight-cut-off (MWCO) Slide-A-Lyzer (Thermo Fisher) for 72 hours at 4°C, against a 50-fold excess of aCSF buffer, which was changed every 12 hours. These samples were aliquoted into small volumes and stored at -80°C, and used for further experiments with no further freeze-thaw cycles.

Materials for SiMPull assays

Biotinylated 6E10 (Biolegend, Cat. No. 803007, Lot No. B230416), Alexa-Fluor-647-labeled 6E10 (Biolegend, Cat. No. 803021, Lot No. B304121), Alexa-Fluor-488-labeled anti-apoE antibody F-9 (Santa Cruz Biology, Cat. No. sc-390925 AF488, Lot No. B1717), Alexa-Fluor-647-labeled Mouse IgG1 (isotype control, MGL, Cat. No. M075-A64, Lot. No. 002), Alexa-Fluor-488-labeled Mouse IgG (isotype control, Fisher Scientific, Cat. No. 65-0865-14) antibodies were used without further modification or purification and stored at 4°C until needed. Anti-apoE antibody EPR19392 (Abcam, Cat. No. ab227993, Lot No. GR3268327-5,130 µg, at 1.3 mg/mL in PBS) was mixed with twenty molar equivalents of Alexa Fluor 488 TFP ester (17.2 nmol, 8 mM in dimethyl sulfoxide) and 10 µL of 1 M NaHCO₃, and incubated for 2 h. The excess dye was removed using a Zeba Spin Desalting Column (7 kDa MWCO), then an Amicon spin filter (50 kDa MWCO), and the concentration of antibody and degree of labeling (~ 1.5 dyes per antibody) were measured using A495 and A280 (Nanodrop, Thermo Fisher). The labeled antibody was diluted to 0.5 mg/mL in 0.5 x PBS containing 0.01% sodium azide and 50% glycerol and stored at -20°C until needed. NeutrAvidin (Thermo Fisher, Cat. No. 31000) and bovine serum albumin (BSA, Thermo Fisher, Cat. No. 10829410) were used without further purification and stored at 4°C and − 20°C, respectively. Glass coverslips (26x76 mm, thickness #1.5, VWR, Cat. No. MENZBC026076AC40) were covalently PEGylated precisely as we reported previously and stored in a desiccator at -20°C until needed.

SiMPull protocol

Assay wells were first coated with 10 µL of 0.2 mg/mL NeutrAvidin in 1× PBS containing 0.05% Tween-20 for 5 min, washed twice with 10 µL of 1× PBS containing 0.05% Tween-20, and once with 10 µL of 1× PBS containing 1% Tween-20. Ten microliters of biotinylated 6E10 (10 nM in 1× PBS containing 0.1 mg/mL BSA) were added to each well for 15 min. The wells were then washed twice with 10 µL of 1× PBS containing 0.05% Tween-20, and once with 10 µL of 1× PBS containing 1% Tween-20. The (co-)aggregates taken at different time points (diluted to 1 µM monomer equivalents in PBS) or soaked-brain extracts (neat) were then added and incubated for 1 hour. The wells were then washed twice with 10 µL of 1× PBS containing 0.05% Tween-20, and once with 10 µL of 1× PBS containing 1% Tween-20. The mixture of imaging antibodies (1 nM Alexa-Fluor-488-labeled EPR19392 and 500 pM Alexa-Fluor-647-labeled 6E10 in 1× PBS containing 0.1 mg/mL BSA) were then added and incubated for 30 min. The wells were then washed twice with 10 µL of 1× PBS containing 0.05% Tween-20, and once with 10 µL of 1× PBS containing 1% Tween-20. Finally, 3 µL of 1× PBS were added to each well and samples were sealed using a second plasma-cleaned coverslip.

Imaging setup for SiMPull

Samples were imaged using a custom-built total internal reflection fluorescence (TIRF) microscope based on a Nikon Ti2 microscope fitted with a Perfect Focus unit. Laser beams (488 nm and 635 nm, Oxxius) were circularly polarized using a quarter-wave plate (WPQ05M-405, Thorlabs), then focused onto the back focal plane of a 100x Plan Apo TIRF, NA 1.49 oil-immersion objective lens (Nikon). Fluorescence emission was collected by the same objective and separated from the excitation beam using a dichroic beamsplitter (Di01-R405/488/561/635, Laser 2000). The emitted light was passed through a 488 nm long-pass filter (BLP01-488R, Laser 2000) and bandpass filter (FF01-520/44 − 25, Laser 2000) for Alexa-Fluor-488 emission, and a 635 nm long-pass filter (BLP01-635R-25, Laser 2000) for Alexa-Fluor-647 emission and imaged on an EMCCD camera (Photometrics Evolve). The open-source software Micro-Manager 1.4 was used to automate image acquisition⁶³. Images were acquired in an unbiased way at laser powers of ~ 10 W/cm² and averaged over 50 frames with an exposure time of 50 ms each.

Analysis of SiMPull data

The averaged images acquired with excitation at 488 nm (apolipoprotein E detection channel) and 635 nm (amyloid-beta detection channel) were passed to the Fiji plugin ComDet3⁶⁴ with a self-written python automation code. Particles are differentiated from noise by selecting 20 times intensity significance (20x SD) compared to the background. For spots in two different channels to be considered colocalized, the displacement between their centers of mass (determined by Gaussian fitting) was required to be ≤ 3 pixels. Colocalization by aggregate number was defined as the ratio between the number of colocalized spots and the total number of spots in the amyloid-beta channel. Colocalization by intensity was defined as the ratio between the sum of the integrated intensities of colocalized spots and all spots in the amyloid-beta channel. Random colocalization was estimated by performing the colocalization algorithm with one of the images (50 µm x 50 µm) rotated by 90 degrees, which was less than 1% for all data sets.

BV-2 Microglia Culture and immunohistochemistry: BV-2 cells were maintained in Dulbecco’s modified Eagles medium (DMEM) Glutamax (Thermo Fisher, Cat. No. 10569010), supplemented with 10% fetal bovine serum (FBS) (Thermo Fisher, Cat. No. 26140087), 100 U/mL penicillin and 100 µg/mL streptomycin (P/S) (Thermo Fisher, Cat. No. 15140148) and incubated at 37°C in 5% CO₂ and 95% air. Cells were grown to ~ 80% confluency and sub-cultured in every 2 days. Old medium was removed from the flask, and cells were washed with PBS. The semi-adherent cells were lifted using accutase for 5 minutes at 37°C. After quenching with PBS, the cell suspension was centrifuged at 200 g for 4 minutes. The supernatant was removed, and the cell pellet resuspended in fresh media. The fresh cell suspension was transferred into a new flask with fresh media at the appropriate splitting ratio (1:5, 1:10) and cultured again in DMEM Glutamax with 5% FBS and P/S. Then BV-2 cells were plated onto 96-well poly-D-lysine coated clear flat-bottom plates (Perkin Elmer, Cat. No. 6055500) and left for 48 h. These cells were then treated with Aβ (co-)aggregates for 24 hours. After 24 hours, the medium was collected, snap-frozen in liquid nitrogen, and stored at -80°C for further TNF-α measurement. BV-2 cells were then fixed using 4% paraformaldehyde (PFA) for 20 minutes and permeabilized using 0.2% Triton X-100 for 15 minutes at room temperature. These BV-2 microglia cells were then washed twice with PBS and blocked for 1 hour at room temperature in blocking solution (3% horse serum in PBS containing 0.05% Triton X). Then 50 nM Alexa-Fluor-488-labeled EPR19392 and 25 nM Alexa-Fluor-647-labeled 6E10 were prepared in blocking solution and applied to cells and stored overnight at 4°C. Cells were washed with 3x PBS and stored in PBS until imaging. Fixed cells were imaged using the Opera Phenix high content imaging system (Perkin Elmer) using a 40x water objective (NA 1.1).

Astrocyte Culture and immunohistochemistry

Primary human astrocytes were purchased from ScienCell™ Research Laboratories (Carlsbad, CA, US). Astrocytes were cultured at 37°C in 5% CO₂ and 95% air, in phenol red-free Human Astrocyte Media, (ScienCell™, 1801prf) supplemented with fetal bovine serum (FBS) (ScienCell™,at 0025), penicillin/streptomycin (ScienCell™, Cat. No. 0503) and astrocyte growth supplement (ScienCell™, Cat. No. 1852). Cells were passaged when approaching ~ 85% confluency and were only used up to passage 7. Astrocytes were plated at appropriate densities on black bottom 96-well plates (Perkin Elmer, Cat. No. 6055500) and cultured for at least 24 hours prior to commencing cell treatments. Astrocytes were treated with Aβ (co-)aggregates for 24 hours. After 24 hours, the medium was collected, snap-frozen in liquid nitrogen, and stored at -80°C for further analysis. The cells were washed with 1xPBS and fixed using 4% PFA for 10 minutes at room temperature. Following fixation, cells were washed three times with 1x PBS and permeabilized in 0.3% triton-X 100 (Sigma, Cat. No. No. T8787) in PBS for 5 minutes at ambient temperature. Cells were blocked for 1 hour in 3% BSA-PBS and after that washed three times with PBS and incubated overnight with a mixture of 50 nM Alexa-Fluor-488-labeled EPR19392 and 25 nM Alexa-Fluor-647-labeled 6E10 in PBS with 1% BSA and 0.1 mg/mL horse serum. These astrocytes were then rewashed three times with PBS and imaged in epifluorescence mode using the Opera Phenix High Content Imaging System (PerkinElmer®).

Data analysis of cellular uptake assay: Fifteen fields of view were imaged per well, with 3 or 4 wells per condition. For each field of view, images were taken in three channels: white differential image contrast (DIC) illumination to visualize the cells, 488 nm excitation (detecting apolipoprotein E antibody, Alexa-Fluor-488-labeled EPR19392) and 635 nm excitation (for Aβ42 antibody, Alexa-Fluor-647-labeled 6E10). First, DIC images were used to define cell perimeters, using a mask created around the border of the cells using an automated python-based program. The aim is to distinguish cell and non-cell areas based on the white light images, then apply the mask to the other two channels and measure the respective fluorescence intensities. Firstly, in order to automate the classification process, a model was applied to the program to learn how to distinguish the cell and non-cell areas. Ten white light images were manually labeled using the Trainable Weka Segmentation⁶⁵ in the Fiji interface⁶⁴ and three different class labels were created: background, cell area, and cluster area (where the cells were clustered) to generate a classifier model using these labeled images. A python code applied this model to all the white light images and generated probability maps for each class. Masks for cell areas were then created by only keeping the coordinates with a probability greater than 67% for the cell. Next, these masks were used in the other two fluorescence channels to extract the respective fluorescence intensities from the other two channels within each cell.

ELISA to measure TNF-α in cell media

To determine TNF-α secretion by BV-2 microglial cells and human astrocytes, cell media were collected after incubation of cells with different aggregates over 24 h and stored at -80°C until analyzed. TNF-α in the cell media was measured using the Duoset® enzyme-linked immunosorbent assay (ELISA) development system (R&D Systems, Abingdon, Oxfordshire, UK), according to the manufacturer's protocol.

Membrane permeabilization assay: The membrane permeabilization assays were performed as previously described⁶⁶. Briefly, vesicles (mean diameter of 200 nm) were prepared by extrusion and five freeze-thaw cycles of a 100:1 mixture of phospholipids 16:0–18:1PC (Avanti Lipids, Cat. No. 850457C) and biotinylated lipids 18:1–12:0 Biotin PC (Avanti Lipids, Cat. No. 860563C). The lipid mixture was hydrated in 100 µM Cal-520 dye (Stratech, Cat. No. 21141) dissolved in HEPES buffer (50 mM, pH 6.5). Non-incorporated Cal-520 dye was separated from the dye-filled vesicles using size-exclusion chromatography, using a SuperdexTM 200 Increase 10/300 GL column attached to an AKTA pure system (GE Life Sciences) with a flow rate of 0.5 mL/min. Dye-filled vesicles were immobilized on coverslips (VWR International, 22x22 mm, 631 − 0122), which were first cleaned using an argon plasma cleaner (PDC-002, Harrick Plasma) for ~ 45 minutes to remove any organic and fluorescent impurities. Frame-seal incubation chambers (9x9mm, Biorad, SLF-0601) were then affixed to the coverslips, and 50 µL of a mixture of 100:1 PLL-g-PEG (20 kDa PLL grafted with 2 kDa PEG and 3.5 Lys units/PEG Chains, SuSoS AG) and PLL-g-PEG biotin (20 kDa PLL grafted with 2 kDa PEG and 50% 3.4 kDa PEG-Biotin, SuSoS AG) at ~ 1 mg/mL was added to the coverslips and incubated for ~ 30 min. The coverslips were then washed three times with HEPES, and 50 µL of 0.1 mg/mL NeutrAvidin (Thermo Fisher, Cat. No. 31000) were added and incubated for 15 min. The coverslips were then rewashed three times with HEPES buffer, and 50 µL of the purified vesicles were added to the coverslips. Before imaging, 50 µL of Ca²⁺-containing buffer (phenol red-free Leibovitz's L-15 Medium, Thermo Fisher Cat No. 21083027) was added, and blank images (F_blank) were recorded. The immobilized vesicles were then incubated with 50 µL of the aggregation mixture (100 nM in monomer equivalents) for 20 min and then reimaged (F_sample). Finally, 1 mg/mL of ionomycin (Cambridge bioscience, Cat. No. 1565-5) - which leads to Ca²⁺ saturation inside the vesicles - was added, and the same areas were reimaged (F_ionomycin). The relative influx of Ca²⁺ into an individual vesicle due to protein aggregates was then determined using the following equation: Ca²⁺ influx = F_sample - F_blank/F_ionomycin - F_blank. The average degree of Ca²⁺ influx was calculated by averaging the Ca²⁺ influx in > 200 individual vesicles. For the imaging, the Cal-520 dye was excited at 488 nm, and the emission from the dye was passed through a combination of a long-pass filter (BLP01-488R-25, Laser 2000) and a bandpass filter (FF01-520/44 − 25, Laser 2000) before being imaged using a Photometrics Evolve EMCCD camera. Images were acquired with a ~ 10 W/cm² power density with a scan speed of 20 Hz and bit depth of 16 bits.

LDH cytotoxicity assay: The assay was performed using an LDH Cytotoxicity Assay Kit (Thermo Fisher, Cat. No. C20303) using human neuroblastoma SH-SY5Y (ATCC) cells cultured in DMEM media without Phenol Red (Gibco, Thermo Fisher, Cat. No. 21063029) and supplemented with 10% Fetal Bovine Serum (Gibco, Thermo Fisher, Cat. No. A4766801). Cells were then treated with aliquots of the aggregation reactions at t₁ and t₃ time points (final Aβ42 concentration: 2 µM in monomer equivalents) for 12 hours, and the cell supernatant was assayed for LDH. The supernatant from cells treated with RIPA lysis buffer (Thermo Fisher, Cat. No. 89900) was used as a positive control, whereas media of the untreated SH-SY5Y cells were taken as a negative control. Next, the supplied reaction mixture (100 µL), which contains a substrate for the LDH activity detection reagent, was added to the medium according to the manufacturer's protocol. The reactions were quenched after 30 minutes with the Stop buffer (also supplied with the kit), and the absorbance at 480 nm was measured using a plate reader (Clariostar -BMG Biotech).

Immunoprecipitation assay

Twenty-fiive microliters of Dynabeads Protein G (Invitrogen, Cat. No. 10007D) were mixed with 100 µL of a 100 nM solution of either HAE-4¹⁵ or Mouse Isotype IgG2a kappa monoclonal [MG2a-53] (Biolegend) antibody in PBS and mixed in a tube rotator for 30 min at room temperature. The beads were then separated using a magnetic separation rack (DynaMag™-2 Magnet, Thermo Fisher, Cat. No. 12321D) and washed twice with PBS. Then, 300 µL of brain extract from APOE4/4 homozygous AD patients or a mixture of in vitro-prepared lipidated and non-lipidated apoE-Aβ (150 µL each), from time-point t₁ were combined with the beads. This mixture was then rotated at 4°C for 3 h. After that, the magnetic rack was used to separate the beads and collect the supernatant for imaging and cellular assays.

Statistical tests

Data are displayed as the mean and error bars represent the standard deviation across replicates respectively. To assess the statistical significance of the differences between two populations we used unpaired two-sample t-tests. Multiple groups were compared using the one-way ANOVAs with post hoc Tukey-tests. Values of p < 0.05 were considered statistically significant. The number of independent biological replicates and technical replicates for each experiment as well as details of statistical tests for each experiment can be found in figure legends. Statistical tests were calculated with Origin 9.0 (OriginLab).

Code availability

Custom scripts used to run the cell uptake and colocalization analyses described in this study are available vis Github.

https://github.com/zengjiexia/CellIntakeAnalysis

https://github.com/zengjiexia/CoincidenceAnalysis

Data availability

All data are available from the corresponding authors upon reasonable request.

Liu, C. C., Kanekiyo, T., Xu, H. & Bu, G. Apolipoprotein E and Alzheimer disease: risk, mechanisms and therapy. Nat. Rev. Neurol. 9, 106–18 (2013).
Conejero-Goldberg, C. et al. APOE2 enhances neuroprotection against Alzheimer’s disease through multiple molecular mechanisms. Mol. Psychiatry 19, 1243–1250 (2014).
Corder, E. H. et al. Gene dose of apolipoprotein E type 4 allele and the risk of Alzheimer’s disease in late onset families. Science (80-.). 261, 921–923 (1993).
Verghese, P. B., Castellano, J. M. & Holtzman, D. M. Apolipoprotein E in Alzheimer’s disease and other neurological disorders. Lancet Neurol. 10, 241–252 (2011).
Kanekiyo, T., Xu, H. & Bu, G. ApoE and Aβ; in Alzheimer’s Disease: Accidental Encounters or Partners? Neuron 81, 740–754 (2014).
Hardy, J. & Selkoe, D. J. The Amyloid Hypothesis of Alzheimer’s Disease: Progress and Problems on the Road to Therapeutics. Science (80-.). 297, 353–356 (2002).
Benilova, I., Karran, E. & De Strooper, B. The toxic Aβ oligomer and Alzheimer’s disease: an emperor in need of clothes. Nat. Neurosci. 15, 349–57 (2012).
Li, S. et al. Soluble oligomers of amyloid β protein facilitate hippocampal long-term depression by disrupting neuronal glutamate uptake. Neuron 62, (2009).
De Strooper, B. et al. Deficiency of presenilin-1 inhibits the normal cleavage of amyloid precursor protein. Nature 391, 387–390 (1998).
Masters, C. L. et al. Amyloid plaque core protein in Alzheimer disease and Down syndrome. Proc. Natl. Acad. Sci. 82, 4245–4249 (1985).
Choi, S. H. et al. A three-dimensional human neural cell culture model of Alzheimer’s disease. Nature 515, 274–278 (2014).
Gonneaud, J. et al. Relative effect of APO ε4 on neuroimaging biomarker changes across the lifespan. Neurology 87, 1696–1703 (2016).
Fleisher, A. S. et al. Apolipoprotein E ε4 and age effects on florbetapir positron emission tomography in healthy aging and Alzheimer disease. Neurobiol. Aging 34, 1–12 (2013).
Liu, C.-C. et al. ApoE4 Accelerates Early Seeding of Amyloid Pathology. Neuron 96, 1024–1032.e3 (2017).
Liao, F. et al. Targeting of nonlipidated, aggregated apoE with antibodies inhibits amyloid accumulation. J. Clin. Invest. 128, 2144–2155 (2018).
Wahrle, S. E. et al. ABCA1 Is Required for Normal Central Nervous System ApoE Levels and for Lipidation of Astrocyte-secreted apoE*. J. Biol. Chem. 279, 40987–40993 (2004).
Mouchard, A. et al. ApoE-fragment/Aβ heteromers in the brain of patients with Alzheimer’s disease. Sci. Rep. 9, 3989 (2019).
Mormino, E. C. et al. Amyloid and APOE ε4 interact to influence short-term decline in preclinical Alzheimer disease. Neurology 82, 1760–1767 (2014).
Lim, Y. Y. et al. APOE and BDNF polymorphisms moderate amyloid β-related cognitive decline in preclinical Alzheimer’s disease. Mol. Psychiatry 20, 1322–1328 (2015).
Kantarci, K. et al. APOE modifies the association between Aβ load and cognition in cognitively normal older adults. Neurology 78, 232–240 (2012).
Castellano, J. M. et al. Human apoE Isoforms Differentially Regulate Brain Amyloid-β Peptide Clearance. Sci. Transl. Med. 3, 89ra57 (2011).
Xiong, M. et al. APOE immunotherapy reduces cerebral amyloid angiopathy and amyloid plaques while improving cerebrovascular function. Sci. Transl. Med. 13, eabd7522 (2021).
Näslund, J. et al. Characterization of stable complexes involving apolipoprotein E and the amyloid β peptide in Alzheimer’s disease brain. Neuron 15, 219–228 (1995).
Huynh, T.-P. V, Davis, A. A., Ulrich, J. D. & Holtzman, D. M. Apolipoprotein E and Alzheimer’s disease: the influence of apolipoprotein E on amyloid-β and other amyloidogenic proteins: Thematic Review Series: ApoE and Lipid Homeostasis in Alzheimer’s Disease. J. Lipid Res. 58, 824–836 (2017).
Gratuze, M. et al. APOE Antibody Inhibits Aβ-Associated Tau Seeding and Spreading in a Mouse Model. Ann. Neurol. 91, 847–852 (2022).
Verghese, P. B. et al. ApoE influences amyloid-β (Aβ) clearance despite minimal apoE/Aβ association in physiological conditions. Proc. Natl. Acad. Sci. 110, 1807–1816 (2013).
Deane, R. et al. apoE isoform–specific disruption of amyloid β peptide clearance from mouse brain. J. Clin. Invest. 118, 4002–4013 (2008).
Ono, K., Condron, M. M. & Teplow, D. B. Structure–neurotoxicity relationships of amyloid β-protein oligomers. Proc. Natl. Acad. Sci. 106, 14745–14750 (2009).
De, S. et al. Different soluble aggregates of Aβ42 can give rise to cellular toxicity through different mechanisms. Nat. Commun. 10, 1541 (2019).
Xian, X. et al. Reversal of ApoE4-induced recycling block as a novel prevention approach for Alzheimer’s disease. Elife 7, e40048 (2018).
Minta, K. et al. Quantification of total apolipoprotein E and its isoforms in cerebrospinal fluid from patients with neurodegenerative diseases. Alzheimers. Res. Ther. 12, 19 (2020).
Galvagnion, C. et al. Lipid vesicles trigger α-synuclein aggregation by stimulating primary nucleation. Nat. Chem. Biol. 11, 229–234 (2015).
Tahirbegi, B. et al. A Novel Aβ40 Assembly at Physiological Concentration. Sci. Rep. 10, 9477 (2020).
Lindberg, D. J., Wesén, E., Björkeroth, J., Rocha, S. & Esbjörner, E. K. Lipid membranes catalyse the fibril formation of the amyloid-β (1–42) peptide through lipid-fibril interactions that reinforce secondary pathways. Biochim. Biophys. Acta - Biomembr. 1859, 1921–1929 (2017).
Jain, A. et al. Probing cellular protein complexes using single-molecule pull-down. Nature 473, 484–488 (2011).
Cerf, E., Gustot, A., Goormaghtigh, E., Ruysschaert, J.-M. & Raussens, V. High ability of apolipoprotein E4 to stabilize amyloid-β peptide oligomers, the pathological entities responsible for Alzheimer’s disease. FASEB J. 25, 1585–1595 (2011).
Küffner, A. M. et al. Sequestration within biomolecular condensates inhibits Aβ-42 amyloid formation. Chem. Sci. 12, 4373–4382 (2021).
Hong, W. et al. Diffusible, highly bioactive oligomers represent a critical minority of soluble Aβ in Alzheimer’s disease brain. Acta Neuropathol. 136, 19–40 (2018).
Sideris, D. I. et al. Soluble amyloid beta-containing aggregates are present throughout the brain at early stages of Alzheimer’s disease. Brain Commun. 3, fcab147 (2021).
Koffie, R. M. et al. Apolipoprotein E4 effects in Alzheimer’s disease are mediated by synaptotoxic oligomeric amyloid-β. Brain 135, 2155–2168 (2012).
Hashimoto, T. et al. Apolipoprotein E, Especially Apolipoprotein E4, Increases the Oligomerization of Amyloid β Peptide. J. Neurosci. 32, 15181 LP – 15192 (2012).
Paresce, D. M., Ghosh, R. N. & Maxfield, F. R. Microglial Cells Internalize Aggregates of the Alzheimer’s Disease Amyloid β-Protein;-Protein Via a Scavenger Receptor. Neuron 17, 553–565 (1996).
Liu, C.-C. et al. Astrocytic LRP1 Mediates Brain Aβ Clearance and Impacts Amyloid Deposition. J. Neurosci. 37, 4023 LP – 4031 (2017).
Hickman, S. E., Allison, E. K. & El Khoury, J. Microglial Dysfunction and Defective β-Amyloid Clearance Pathways in Aging Alzheimer’s Disease Mice. J. Neurosci. 28, 8354 LP – 8360 (2008).
Garwood, C. J., Pooler, A. M., Atherton, J., Hanger, D. P. & Noble, W. Astrocytes are important mediators of Aβ-induced neurotoxicity and tau phosphorylation in primary culture. Cell Death Dis. 2, e167–e167 (2011).
Craft, J. M., Watterson, D. M. & Van Eldik, L. J. Human amyloid β-induced neuroinflammation is an early event in neurodegeneration. Glia 53, 484–490 (2006).
Flagmeier, P. et al. Direct measurement of lipid membrane disruption connects kinetics and toxicity of Aβ42 aggregation. Nat. Struct. Mol. Biol. 10, 886–891 (2020).
Hsiao, K. et al. Correlative Memory Deficits, Aβ Elevation, and Amyloid Plaques in Transgenic Mice. Science (80-.). 274, 99–103 (1996).
McLean, C. A. et al. Soluble pool of Aβ amyloid as a determinant of severity of neurodegeneration in Alzheimer’s disease. Ann Neurol 46, 860–866 (1999).
Mandrekar, S. et al. Microglia Mediate the Clearance of Soluble Aβ through Fluid Phase Macropinocytosis. J. Neurosci. 29, 4252 LP – 4262 (2009).
Wang, W.-Y., Tan, M.-S., Yu, J.-T. & Tan, L. Role of pro-inflammatory cytokines released from microglia in Alzheimer’s disease. Ann. Transl. Med. Vol 3, No 10 (June 2015) Ann. Transl. Med. (2015).
Griffin, W. S. T. et al. Glial-Neuronal Interactions in Alzheimer’s Disease: The Potential Role of a ‘Cytokine Cycle’ in Disease Progression. Brain Pathol. 8, 65–72 (1998).
Hanson, A. J. et al. Effect of Apolipoprotein E Genotype and Diet on Apolipoprotein E Lipidation and Amyloid Peptides: Randomized Clinical TrialApolipoprotein E Lipidation and Amyloid Peptides. JAMA Neurol. 70, 972–980 (2013).
Hu, J. et al. Opposing effects of viral mediated brain expression of apolipoprotein E2 (apoE2) and apoE4 on apoE lipidation and Aβ metabolism in apoE4-targeted replacement mice. Mol. Neurodegener. 10, 6 (2015).
Wahrle, S. E. et al. Deletion of Abca1 Increases Aβ Deposition in the PDAPP Transgenic Mouse Model of Alzheimer Disease. J. Biol. Chem. 280, 43236–43242 (2005).
Koldamova, R., Staufenbiel, M. & Lefterov, I. Lack of ABCA1 Considerably Decreases Brain ApoE Level and Increases Amyloid Deposition in APP23 Mice. J. Biol. Chem. 280, 43224–43235 (2005).
Wahrle, S. E. et al. Overexpression of ABCA1 reduces amyloid deposition in the PDAPP mouse model of Alzheimer disease. J. Clin. Invest. 118, 671–682 (2008).
Frieden, C., Wang, H. & Ho, C. M. W. A mechanism for lipid binding to apoE and the role of intrinsically disordered regions coupled to domain–domain interactions. Proc. Natl. Acad. Sci. 114, 6292–6297 (2017).
Chen, Y., Strickland, M. R., Soranno, A. & Holtzman, D. M. Apolipoprotein E: Structural Insights and Links to Alzheimer Disease Pathogenesis. Neuron 109, 205–221 (2021).
Holtzman, D. M., Herz, J. & Bu, G. Apolipoprotein E and Apolipoprotein E Receptors: Normal Biology and Roles in Alzheimer Disease. Cold Spring Harb. Perspect. Med. 2, (2012).
Yeh, F. L., Wang, Y., Tom, I., Gonzalez, L. C. & Sheng, M. TREM2 Binds to Apolipoproteins, Including APOE and CLU/APOJ, and Thereby Facilitates Uptake of Amyloid-Beta by Microglia. Neuron 91, 328–340 (2016).
Narayan, P. et al. PICALM Rescues Endocytic Defects Caused by the Alzheimer’s Disease Risk Factor APOE4. Cell Rep. 33, 108224 (2020).
Edelstein, A. D. et al. Advanced methods of microscope control using µManager software. J. Biol. Methods 1, 1–10 (2014).
Schindelin, J. et al. Fiji: an open-source platform for biological-image analysis. Nat. Methods 9, 676–82 (2012).
Arganda-Carreras, I. et al. Trainable Weka Segmentation: a machine learning tool for microscopy pixel classification. Bioinformatics 33, 2424–2426 (2017).
Flagmeier, P. et al. Ultrasensitive measurement of Ca2 + influx into lipid vesicles induced by protein aggregates. Angew. Chemie Int. Ed. 56, 7750–7754 (2017).

Author Contributions

Z.X, E.E.P., R.T.R. and S.D. did the SiMPull experiments. R.T.R. and E.E.P. modified antibodies and coated coverslips for SiMPull. Z.X., H.E.W and S.D. aggregated proteins. The cellular uptake assay was carried out by M.M., E.E.P and H.E.W. A.H. helped with the uptake data analysis. H.D. and S.D. prepared soaked-brain extracts. Z.X., E.E.P., and S.D. did the membrane permeabilization and LDH assays. H.J., H.Z., Y.L.J.L., Y.Z., and J.D. helped with antibody modifications, slide preparation or microscopy. Data analysis and statistics were developed, carried out by Z.X., and interpreted by Z.X. and S.D. L.F., S.W., S.M.B., R.T.R., A.H., H.M., P.T., D.C.C., D.M.H., D.K., and S.D. supervised the project. The initial draft of the paper was written by R.T.R., D.K., E.E.P. and S.D.; all other authors provided feedback and contributed in editing the manuscript into its final form. D.K. and S.D. designed and conceived the study. All authors read and approved the manuscript.

Acknowledgments

The tissue for this study was provided by the Newcastle Brain Tissue Resource, which is funded in part by a grant from the UK Medical Research Council and in part by Brains for Dementia Research, a joint venture between Alzheimer's Society and Alzheimer's Research UK. We also thank the Professor Maria Spillantini laboratory for providing the facilities and expertise for processing of the brain tissue used in these experiments.

Funding

This study was supported by the Parkinson’s UK grants F-1301 and F-1301 (H.M.), NIH grants AG047644 (D.M.H.) and NS090934 (D.M.H.), European Research Council Grant Number 669237 (D.K.), the Royal Society (D.K.), Dementia Research UK Pilot Award (S.D.) and the UK Dementia Research Institute (DRI) at Cambridge (D.K.) and a UKRI Future Leaders Fellowship (Grant number MR/V023861/1) (S.D.). The Sheffield NIHR Biomedical Research Centre provided support for this study.

Competing interests

D.M.H. is an inventor on a patent licensed by Washington University to NextCure on anti-apoE antibodies. D.M.H. co-founded and is on the scientific advisory board of C2N Diagnostics, DenaliGenentech, and Cajal Neuroscience. D.M.H. consults for Alector. The lab of D.M.H. receives research grants from the National Institutes of Health, Cure Alzheimer’s Fund, Tau Consortium, the JPB Foundation, Good Ventures, the Rainwater Foundation, NextCure, Denali, and Ionis. D.C.C. and P. T. hold stock in AstraZeneca. All the other authors declare no conflicts of interest.

Yes there is potential Competing Interest. D.M.H. is an inventor on a patent licensed by Washington University to NextCure on anti-apoE antibodies. D.M.H. co-founded and is on the scientific advisory board of C2N Diagnostics, DenaliGenentech, and Cajal Neuroscience. D.M.H. consults for Alector. The lab of D.M.H. receives research grants from the National Institutes of Health, Cure Alzheimer’s Fund, Tau Consortium, the JPB Foundation, Good Ventures, the Rainwater Foundation, NextCure, Denali, and Ionis. D.C.C. and P. T. hold stock in AstraZeneca. All the other authors declare no conflicts of interest.

XiaetalSI.docx

Download PDF

Journal Publication

published 31 May, 2024

Read the published version in Nature Communications →

Version 1

posted

You are reading this latest preprint version

Co-aggregation with Apolipoprotein E modulates the function of Amyloid-β in Alzheimer's disease

Status:

Journal Publication

Version 1

Abstract

Figures

Introduction

Results

Soluble apoE-Aβ Co-aggregates Form Isoform-independently, but Accumulate Isoform-dependently in AD Brains

Glial Uptake of Soluble apoE-Aβ Co-aggregates is Isoform- and Lipidation-dependent

Soluble, Non-lipidated apoE4-Aβ Co-aggregates are Toxic and Impair Clearance by Glial Cells

Soluble apoE4-Aβ Co-aggregates Detected in AD Brains are Poorly Lipidated

Discussion

Methods

Code availability

Data availability

References

Declarations

Additional Declarations

Supplementary Files

Status:

Journal Publication

Version 1