Background
An important breakthrough in our understanding of Alzheimer’s disease (AD) was the identification of the apolipoprotein E
APOE-ɛ4 allele as a risk factor [
1]. Apolipoprotein E (apoE) protein is a component of lipoprotein particles in the plasma, as well as in the cerebrospinal fluid (CSF) [
2]. ApoE regulates important signaling pathways by interacting with receptors and is present as sialylated glycoforms [
3]. Human apoE lacks the consensus sequence necessary for N-linked glycosylation; thus, O-linked carbohydrates probably account for glycosylation [
4]. The impact of apoE glycosylation remains unclear, but evidence indicates that glycosylation acts as an important post-translational mechanism for fine-tuning apoE interaction with receptors and proteins [
5].
In humans, three versions of the
APOE gene exist, ε2 (apoE2), ε3 (apoE3), and ε4 (apoE4) alleles, while other mammals only have one version of the
APOE gene, resembling ancestral apoE4 [
6].
APOE-ε3 is the most common allele (~75%), followed by ε4 (15–20%) and ε2 (4–8%) [
7]. Compared to the most common
APOE ε3/ε3 genotype, each additional copy of the
APOE-ε4 allele is associated with a higher risk of AD and a younger mean age of dementia onset. Thus, in individuals with one copy of the
APOE-ε4 allele, the risk of AD increases 2–3 times and 8–12 times in individuals with two copies [
8]. Experimental evidence shows the deleterious effect of the apoE4 variant for AD, while the lack of apoE4 appears to be protective [
9]. In contrast, the presence of one or two copies of the
APOE-ε2 allele is associated with a lower risk of AD and an older mean age of dementia onset [
10]; therefore, it has been hypothesized that the apoE2 protein could be protective against AD [
11]. Indeed,
APOE-ε2 homozygotes present an exceptionally low likelihood of developing AD [
12]. The reported effects of different
APOE genotypes on AD risk vary widely with demographic factors such as gender and ethnicity [
7]. Moreover, the percentage of
APOE genotypes in cognitively unimpaired people with neuropathological or biomarker evidence of preclinical AD, or the percentage of people who meet the criteria for mild cognitive impairment with or without biomarker evidence of AD, is not well established (discussed in [
12]). Anyhow, despite the 2–3-fold increase in AD prevalence in
APOE-ε4 subjects compared to the general population, most of the individuals with AD are
APOE-ε3 homozygotes [
13].
Nonetheless, given the important physiological functions of apoE, a malfunctioning of the apoE protein may also contribute to AD pathology in ε4 non-carriers [
14]. The differences in the structure of apoE isoforms influence their ability to bind lipids, receptors, and amyloid-β (Aβ), which aggregates in plaques within the brain of AD patients [
14].
Interestingly, apoE forms disulfide-linked homodimers and heterodimers with the apoA-II apolipoprotein involving the cysteine (Cys) at position 112 [
7,
14]. Indeed, these apoE homodimers linked by disulfide bonds could be the native form able to bind to receptors [
15]. The three human apoE isoforms differ in the presence of Cys/arginine (Arg) at positions 112 and 158 within the receptor binding domain, as apoE4 lacks Cys residues at both these positions [
4]. The amino acid substitution of Cys-112 by Arg in apoE4 explains the lower number of disulfide-linked dimers in the CSF of
APOE ε3/ε4 subjects compared with
APOE ε3/3 subjects, and their absence in
APOE ε4/ε4 subjects [
16,
17], but may also explain the reduced ability of apoE4 to mediate some of its biological roles, compared with apoE2 or apoE3 [
18].
The mature apoE protein has 299 amino acids and a molecular mass of ~35 kDa. However, previous studies performed in the brain [
19] and CSF [
17] reported a ~100-kDa apoE band in non-reducing conditions, as opposed to the predicted ~70 kDa, which was referred to as an apoE homodimer.
Previous studies that considered total CSF apoE levels failed to demonstrate consistent changes when the
APOE genotype was included as a covariate in the models [
20‐
22]. However, other studies associated high CSF apoE concentrations with an increased risk of impaired cognitive progression in non-apoE4 carriers [
23].
Anyhow, in order to consider the estimation of apoE levels in CSF as a read-out of AD occurrence or progression, in addition to the APOE genotype, the studies should also consider changes in the protein conformation/structure that can compromise the biological function of the apoE protein. In this study, we aimed to characterize the occurrence of different apoE species in AD CSF from individuals with different APOE genotypes, while considering changes in the balance of apoE glycoforms and the occurrence of aberrant apoE dimers that could indicate a compromise of apoE function in the brain.
Materials and methods
Patients
CSF samples from individuals with known APOE genotypes were obtained from two independent cohorts. The CSF samples from both cohorts used for this study were de-identified aliquots from clinical routine analyses, following procedures approved by the Ethics Committees at the University of Gothenburg and the Hospital Sant Pau, respectively. Additionally, this study was approved by the ethics committee at the Miguel Hernandez University, and was carried out in accordance with the Helsinki Declaration regarding research on humans.
The CSF samples were obtained by lumbar puncture and centrifuged (2000×g, 10 min) and then immediately aliquoted and stored in ultrafreezers and kept at −80°C until analysis. The time between CSF acquisition and storage was less than 4 h in all cases. The handling of the samples was performed following recommended operating procedures [
24]. Freeze-thaw cycles were avoided and new aliquots were used for each independent analysis.
The first cohort was from the longitudinal geriatric population study in Piteå, Sweden [
25], the Piteå Dementia Project. The diagnostic evaluation included a clinical examination (detailed medical history and somatic, neuropsychiatric, and neurological status), a neuropsychological test battery, routine blood and CSF tests, and a CT scan to exclude secondary dementias [
26]. All clinical diagnoses and evaluations were made without knowledge of the results of the biochemical analyses and vice versa. The cohort consisted of 45 patients with AD (fourteen men and thirty-one women, mean age 77±1 years) and was selected based on the
APOE-ε4 status, so that fifteen each had
APOE ε3/ε3,
APOE ε3/ε4, or
APOE ε4/ε4. In addition, fourteen non-AD controls [seven men and seven women, mean age (67 ± 3 years);
APOE ε3/ε3: 9,
APOE ε3/ε4: 5] were included.
APOE genotype was determined by the solid-phase mini-sequencing method as previously described [
27]. For this study, patients who were designated as AD or controls also had typical core CSF biomarker levels [Aβ42 and total tau (T-tau)] using cut-offs that are >90% specific for AD [
28], but except for CSF Aβ42 and T-tau, all biochemical analyses were made without knowledge of the clinical data. The ethics committees in Umeå University and University of Gothenburg approved the study.
The second cohort was obtained from the Sant Pau Initiative on Neurodegeneration (SPIN cohort) [
29] from Hospital Sant Pau (Barcelona, Spain). We included samples from 29 AD patients (thirteen men and sixteen women, mean age 73±1 years;
APOE: 10 ε3/ε3, 10 ε3/ε4, 9 ε4/ε4) and ten controls (seven men and three women, mean age 69±2 years;
APOE: 5 ε3/ε3, 5 ε3/ε4). Typically, these are patients who present cognitive complaints and are referred to the specialized memory unit from their primary care physician. All patients undergo a full neuropsychological evaluation that demonstrates objective cognitive impairment. Patients were included in the cohort when they presented supportive biomarkers of the AD pathophysiological process. Cognitively normal participants were volunteers without cognitive complaints and normal neuropsychological evaluation. More details about inclusion/exclusion criteria and neuropsychological tests in this cohort are detailed elsewhere [
29].
In this cohort, the
APOE genotype was determined by direct DNA sequencing and visual analysis of the resulting electropherogram performed to identify the two coding polymorphisms that encode the three possible apoE variants [
29].
Each center applied their own internally validated cut-offs, according to their preanalytical and analytical particularities. More details about the cut-offs applied are indicated below. Samples were retrospectively selected from large cohorts to balance age, sex, and
APOE status. Most of the selected cases (92%, 43 of 45 from Gothenburg and 25 of 29 from Barcelona) were categorized A+T+ according to [
30]; thus, subgrouping by the AT(N) system for analysis was impractical. For full details about the collections, see Table
1.
Table 1
Demographic and biomarker information from the CSF samples obtained from the Gothenburg (Sweden) and Barcelona (Spain) cohorts
Cohort: Gothenburg (Sweden) |
| Control | Alzheimer’s disease |
APOE | ε3/ε3 | ε3/ε4 | All | ε3/ε3 | ε3/ε4 | ε4/ε4 | All |
N | 9 | 5 | 14 | 15 | 15 | 15 | 45 |
Age (years) | 69±2 | 62±5 | 67±3 | 79±2 | 78±1 | 73±1 | 77±1* |
Age (range) | 60–81 | 44–75 | 44–81 | 62–88 | 69–84 | 63–83 | 62–88 |
Female/male | 5/4 | 2/3 | 7/7 | 11/4 | 11/4 | 9/6 | 31/14 |
CSF Aβ42 (pg/mL) | 845±96 | 746±121 | 804±74 | 470±13* | 480±8* | 419±21 | 457±10* |
CSF tau (pg/mL) | 317±53 | 303±34 | 312±35 | 816±88* | 917±112* | 731±53 | 840±52* |
Cohort: Barcelona (Spain) |
| Control | Alzheimer’s disease |
APOE | ε3/ε3 | ε3/ε4 | All | ε3/ε3 | ε3/ε4 | ε4/ε4 | All |
N | 5 | 5 | 10 | 10 | 10 | 9 | 29 |
Age (years) | 71±2 | 67±52 | 69±2 | 75±2 | 73±2 | 72±2 | 73±1* |
Age (range) | 66–76 | 60–72 | 60–76 | 64–84 | 64–83 | 61–85 | 61–85 |
Female/male | 1/4 | 2/3 | 3/7 | 7/3 | 2/8 | 7/2 | 16/13 |
CSF Aβ42 (pg/mL) | 1139±248 | 1010±116 | 1075±131 | 607±60* | 543±37* | 493±62 | 549±31* |
CSF tau (pg/mL) | 295±49 | 261±23 | 278±26 | 778±94* | 624±62* | 908±81 | 765±50* |
Determination of AD core biomarkers by ELISA and definition of cut-offs
In the cohort from Gothenburg, the levels of the AD core biomarkers T-tau, P-tau, and Aβ42 were measured in the CSF using INNOTEST ELISAs (Fujirebio-Europe, Gent, Belgium). Patients were designated as AD or controls according to CSF biomarker levels using cut-offs that are >90% specific for AD: Aβ42 <550 pg/mL and total tau (T-tau) >400 pg/mL [
20].
For the cohort from Barcelona, cut-offs for AD biomarkers measured in the Lumipulse automated platform (Fujirebio-Europe) were T-tau > 400 pg/mL, P-tau > 63 pg/mL, and 0.062 for the Aβ42/Aβ40 ratio [
29].
All samples were analyzed as part of a clinical routine by board-certified laboratory technicians following strict procedures for batch-bridging, analyses, and quality control of individual ELISA plates.
Transgenic rat CSF
The experiments were carried out using a cohort of 107 rats (53 males and 54 females), including transgenic TgF344-AD rats (
n = 52) expressing mutant human APP (APPsw) and presenilin-1 (PS1ΔE9) genes [
31] and wild-type Fischer rats (
n = 55). Rats were bred in the animal research facilities at the University of Barcelona. Animals were provided with food and water ad libitum and maintained in a temperature-controlled environment in a 12/12-h light-dark cycle. CSF samples (50–100 μL) were collected from ketamine/xylazine-anesthetized animals by cisternal puncture with a glass capillary in the suboccipital region through the atlanto-occipital membrane, with a single incision into the subarachnoid space [
32]. CSF aliquots from different time points [4 months: 16 wild-type (8 male, 8 female) and 16 TgF344-AD animals (8 male, 8 female); 10.5 months: 17 wild-type (8 male, 9 female) and 16 TgF344-AD animals (8 male, 8 female); 16.5 months: 22 wild-type (12 male, 10 female) and 20 TgF344-AD animals (9 male, 11 female)] were analyzed. This study was part of a large project assessing various different proteins that included brain analysis at each stage; thus, it was not possible to perform longitudinal measurements in the same animal (repeat sampling) to reduce the number of animals. Animal work was performed in accordance with the local legislation, with the approval of the Experimental Animal Ethical Committee of the University of Barcelona, and in compliance with European legislation.
Western blotting
Samples of human or rat CSF (10 μL) were denatured at 98°C for 5 min and resolved by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) under reducing or non-reducing conditions (determined by the presence or absence of β-mercaptoethanol in the sample buffer, respectively). Unless specified, the studies presented in the text were performed under reducing conditions. For this study, we used 12% precast gels (Bio-Rad Laboratories, GmbH, Munich, Germany; #4561046). All the samples were analyzed at least in duplicate (duplicates in separate gels) and distributed in the gels to ensure the comparison by disease condition and APOE genotype. The distribution of the samples in the gels was performed by a member of the team and the experiments were performed by another, the experimenter, in a blind way.
Following electrophoresis, proteins were blotted onto 0.45-μm nitrocellulose membranes (Bio-Rad Laboratories, GmbH, Munich, Germany). Bands of apoE immunoreactivity were detected using either the antibody AB178479 (goat polyclonal; Merck Millipore) or the antibody AB947 (goat polyclonal; Merck Millipore), both common to all apoE isoforms, or alternatively by an antibody specific to the apoE4 isoform (recognizes an internal domain comprising the Arg112 residue present exclusively in apoE4 species; mouse monoclonal, Novus Biologicals; NBP1-49529). Blots were then probed with the appropriate conjugated secondary antibodies (IRDye secondary antibodies, LI-COR Biosciences, Lincoln, NE, USA) and imaged on an Odyssey CLx Infrared Imaging System (LI-COR Biosciences). Band intensities were analyzed using LI-COR software (ImageStudio Lite). The boxes selected with the ImageQuant Studio software for quantification, as well as the completed blots, are shown as
supplementary figures. Recombinant apoE3 (Peprotech, ThermoFisher Scientific# 350-02) was included into each blot to serve as a loading reference and for normalizing the immunoreactivity signal between blots. Specifically, the same amount of recombinant apoE3 was always included, and the immunoreactivity of the apoE bands from each blot was referred to (divided by) the immunoreactivity of recombinant apoE3, thus correcting inter-blot differences and allowing for comparisons across assays.
For blue-native gel electrophoresis, the CSF samples were not heated (native conditions) and were loaded with NuPage LDS 4× Sample Buffer (ThermoFisher Scientific, NP007) into native-PAGE 4–16% gels (ThermoFisher Scientific, BN1002BOX). Buffers were prepared using native-PAGE Running Buffer (ThermoFisher Scientific, BN2001) and native-PAGE Cathode Buffer Additive (ThermoFisher Scientific, BN2002). Immunoreactivity was detected using the AB178479 antibody and HRP anti-goat secondary antibody (ThermoFisher). The signal was visualized by ECL (GE Healthcare Life Science) and analyzed using ImageStudio Lite.
ApoE immunoprecipitation
CSF samples (50 μL) were incubated on a roller overnight with 100 μL PureProteome FlexiBind Magnetic Beads (Merck Millipore, LSKMAGN04) coupled with the AB178479 apoE antibody (Merck Millipore). The supernatant was removed, and the beads were washed and then resuspended and boiled at 98 °C for 5 min in SDS-PAGE sample buffer and analyzed by western blot with the AB947 antibody (Merck Millipore) or anti-apoE4 antibody (Novus Biologicals, NBP1-49529). For a control immunoprecipitation, beads were coupled with horse serum and then incubated with CSF samples.
Enzymatic deglycosylation
Enzymatic deglycosylation was performed using an Agilent Enzymatic Deglycosylation Kit (Agilent Technologies, GK80110) following the manufacturer’s instructions. Briefly, for each condition, 30 μL of control or AD CSF was mixed with 10-μl incubation buffer and 2.5-μL denaturing buffer and heated at 100 °C for 5 min. The samples were then cooled down to room temperature, and 2.5 μL of detergent (15% NP-40) was added while mixing gently. O- (1 μL sialidase and 1 μL O-glycanase) or N-linked (1 μL N-glycanase) deglycosylating enzymes were then added according to each different condition (O-linked, N-linked, or O- and N-linked deglycosylation) and samples were heated at 37 °C for 3 h. Samples were then analyzed by western blot. As for control of the deglycosylation process, samples exposed to the same heating conditions but without deglycosylating enzymes were included.
In-gel digestion
In-gel digestion was performed as previously described [
33] in order to investigate the content of western blot immunoreactive bands of interest using an antibody-free method. Briefly, 1 mL of a pool of AD CSF (
APOE ε3/ε4 and
APOE ε3/ε4 cases) was immunoprecipitated with AB178479 antibody and loaded into SDS-polyacrylamide gel under reducing conditions, as described above. ApoE3 and apoE4 recombinant proteins (Peprotech, ThermoFisher Scientific# 350-02 and 350-04) were also loaded in the gel (10 pmol) and used as a reference for band excising and positive control. Upon electrophoresis, the gel was divided into two parts, one for protein visualization by SimplyBlue
TM SafeStain Coomassie (ThermoFisher Scientific, cat# LC6060) and one for blotting with the AB947 antibody as confirmation of band presence and location. Bands of interest were cut-out from the AD CSF gel lane and recombinant protein lanes and destained using a 1:1 mixture of acetonitrile and 50 mM ammonium bicarbonate solution twice for 15 min. Furthermore, gel pieces were de-hydrated with 100% acetonitrile and dried using a vacuum centrifuge. Samples were subsequently reduced with 10 mM dithiothreitol (DTT) for 1 h at 56 °C and alkylated with 25 mM iodoacetamide (IAA) for 45 min at room temperature in the dark. Gel pieces were further washed with 25mM ammonium bicarbonate, de-hydrated with 100% acetonitrile, and dried using a vacuum centrifuge once more. Samples were digested overnight at 37°C using 100 ng/μL trypsin enzyme (Sequencing Grade Modified Trypsin, #V511A, Promega). The next day, digestion was stopped by the addition of 2% trifluoroacetic acid and 75% acetonitrile solution, and peptides were collected into a new tube (Costar, #3207). Gel pieces were further extracted with the addition of 50% acetonitrile and 0.2% trifluoroacetic acid solution shaking for 30 min. The supernatant containing the peptides was transferred to the collection tube. Pooled extracts for each gel piece were dried through vacuum centrifugation and stored at −80 °C pending mass spectrometry (MS) analysis.
Mass spectrometry data analysis
Dried in-gel digested samples were reconstituted in 7 μL 8% acetonitrile/8% formic acid solution and shaken for 30 min. A total of 6 μL of each sample was investigated using mass spectrometry (MS) analysis performed with a Dionex 3000 nanoflow liquid chromatography system coupled to a Q Exactive (both Thermo Fisher Scientific). Briefly, a reversed phase Acclaim PepMap C18 (100 Å pore size, 3 μm particle size, 20 mm length, 75 μm i.d., Thermo Fisher Scientific) trap column was used for online desalting and sample clean-up. Separation was performed with a reversed phase Acclaim PepMap RSLC C18 (100 Å pore size, 2 μm particle size, 75 μm i.d., 150 mm length, Thermo Fisher Scientific) column at a flow rate of 300 nL/min by applying a linear gradient of 0–40% B for 50 min. Mobile phase A was 0.1% formic acid in water (v/v) and mobile phase B was 0.1% formic acid and 84% acetonitrile in water (v/v/v).
Mass spectra were acquired in positive ion mode and in a data-dependent manner with a resolution setting of 70,000 for precursor and 17,500 for fragment ion acquisitions. Fragmentation was obtained by higher energy collision-induced dissociation (HCD) using a normalized collision energy (NCE) setting of 28. Database searches were made using PEAKS Studio XPRO (Bioinformatic Solutions, Inc., Waterloo, Canada).
Statistical analysis
All the data was analyzed using GraphPad Prism (version 7; GraphPad software, San Diego, CA, USA). The test was used to analyze the distribution of each variable. Firstly, multiple comparisons were performed between groups, ANOVA was used for parametric variables, and the Kruskal-Wallis test for non-parametric variables. A Student’s t-test for parametric variables and a Mann-Whitney U test for non-parametric variables were employed for comparison between two groups and for determining precise p values. For correlations, the Pearson and Spearman tests were used. The results are shown as means ± SEM; the standard deviation (SD) and median values are also displayed as indicated in the figure legends.
Discussion
Typically, transgenic models produce pathological changes that partially replicate changes seen in human patients. In this study, firstly, we have found an increase in CSF apoE in the TgF344-AD rats, with the documented occurrence of amyloid pathology around 10 months of age [
31,
34]. This result can be interpreted as a suggestive gain of function for apoE in AD. In fact, this increase in CSF apoE content is similar to the one observed in AD patients when considering total apoE content. Considering the summation of the apoE immunoreactivity for 34- and 36-kDa (not including the value for the 100-kDa band) species, in samples from AD patients, a significant overall increase in total CSF apoE was found in the Gothenburg cohort and a non-significant trend to increase was seen in the Barcelona cohort compared to controls. However, the biochemical discrimination of different human CSF apoE species and the altered balance of these species lead us to believe that, despite the increase in total CSF apoE levels determined in the AD transgenic model and AD patients, the imbalance between apoE species should be interpreted as indicative of a potential impairment in apoE function in the brain. Thus, higher levels of apoE could paradoxically result in less functionality if the increase is represented by complexes and immature glycoforms.
Indeed, we have identified two monomeric apoE species in human CSF and demonstrated that the balance between these species in AD patients differs compared to that of controls. Some previous studies that did not distinguish the contribution of particular apoE species have indicated that CSF apoE levels in AD patients are increased [
35], also at follow-up [
36], but many studies addressing total CSF apoE levels are inconclusive and found no clear association with the AD condition or
APOE genotype [
20‐
22]. In addition to recurrent confounding factors such as the handling of the samples, and also considering differences in the diagnostic accuracy between cohorts, the inconsistencies found in these previous reports could be associated mostly with the determination method used, as some are based in MS [
17,
18], while others use immunoassays [
16], both of which fail to discriminate between apoE species. Even if an immunoassay is the most available and desirable approach for quantitative analysis of altered levels of a biomarker, this method does not easily detect subtle changes in specific species (imbalance in glycoforms) and/or does not detect particular species suffering conformational changes (aberrant dimers).
The 34- and 36-kDa species are likely different O-glycoforms, and the difference in electrophoretic mobility of the apoE glycoforms could be a consequence of its sialylation [
37]. ApoE is exclusively O-glycosylated and can be capped with one or two sialic acids [
5]. In CSF, the existence of two glycans per molecule of apoE has been demonstrated [
38], and previous studies indicate that astrocytes secrete two differential glycoforms of apoE [
39] and that the sialo and asialo forms of apoE can both be secreted into the medium [
40]. Our results indicate that the 34-kDa apoE monomers, which appear to be less sialylated than 36-kDa apoE monomers [
3], are present at a higher proportion in AD subjects compared with controls, in both independent cohorts. Whether or not these 34-kDa species can participate in disulfide-linked apoE dimers or pathological complexes, as described here, requires further study.
Moreover, the altered balance between apoE glycoforms should be validated in external cohorts. Here, most of the results obtained in the Gothenburg cohort were validated in a second independent cohort from Barcelona, despite the small size of the groups in this cohort. Nonetheless, some inconsistent results were observed between cohorts regarding the ratio of the 36-kDa/34-kDa species. In the Gothenburg cohort, this ratio was significantly higher in AD individuals with an APOE ε4/ε4 genotype compared with APOE ε3/ε3 and ε3/ε4, while in the Barcelona cohort, the ratios were at a similar level among AD individuals with different APOE genotypes. Additional studies will serve to determine if the imbalance between apoE glycoforms is a common feature for AD APOE-ε4 homozygote subjects.
Indeed, the changes observed in this study are less obvious in ε4/ε4 samples. This discrepancy may be due to the fact that small changes in apoE levels for ε4/ε4 subjects could be more detrimental than in the rest of APOE genotypes, perhaps caused by the basal compromise in some of the biological functions of apoE in the brain related with the inability of the apoE4 isoform to form dimers.
APOE-ε4 is the strongest risk factor gene for AD, although inheriting APOE-ε4 does not mean a person will definitely develop the disease. Thus, the opportunity to analyze the subset of APOE ε3/ε4 control individuals with no AD-like symptoms is very interesting. As stated, all the cases were retrospectively selected from large cohorts and based on the determination of AD core biomarkers. The diagnostic uncertainty is inherent in this type of studies, but the control individuals with APOE ε3/ε4 genotype displayed similar apoE values as the ones obtained in APOE ε3/ε3 individuals.
Correct apoE glycosylation is fundamental for its function and lipoprotein binding capacity. ApoE glycosylation can modulate receptor affinity, lipid-binding ability, lipid transportation, and metabolic functions [
41‐
43]. Furthermore, apoE deglycosylation reduces its binding to Aβ42 [
44] and may induce Aβ42 accumulation [
45]. Our results suggest that the imbalance between the different glycoforms of apoE monomers observed in AD may interfere with its biological function, contributing to the progression of the disease. Interestingly, apoE glycosylation also plays a key role in the protection against self-association and spontaneous aggregation [
46].
As mentioned, the apoE isoforms encoded by
APOE ε3 or ε2 are able to form disulfide-linked hetero- and homodimers through the Cys residue at position 112, while
APOE ε4 (which presents Arg at position 112) and apoE from non-human mammals are unable to form these oligomeric species. However, in our studies, apoE4 isoforms were present in 100-kDa aggregates in
APOE ε3/ε4 AD cases, and these aggregates were identified in most of the
APOE ε4/ε4 AD patients. These 100-kDa complexes are compatible in molecular mass to disulfide-linked apoE dimers, which exist as a major portion of apoE in human CSF of
APOE ε3 or ε2 carriers [
16]. The existence of SDS-resistant dimers of apoE4 was suggested when studying the in vitro formation of SDS-resistant Aβ-apoE complexes [
47]; but, to our knowledge, it has never been demonstrated in vitro or in vivo. The definitive identity of the 100-kDa species was confirmed by the diverse immunoprecipitation analyses combining antibodies originated from diverse animal species and the MS studies. Rats express a unique apoE variant most closely related to the human ε4-type haplotype. However, in the transgenic rat model of AD, we were not able to observe the 100-kDa resistant apoE species that we observed in AD
APOE ε4/4 cases. Likewise, the possibility that inactive monomers of apoE occur in this animal model requires further study; however, models in which the amyloid condition results in an increase of apoE expression should consider this possibility.
ApoE dimers or multimers may be the biologically important species, particularly in receptor binding [
15]. In a previous study, the levels of apoE dimers in the CSF from AD subjects were not different from those in controls [
48], although in this study they did not assess the nature of the aberrant β-mercaptoethanol resistant complexes. In our AD samples, the 100-kDa apoE complexes are aberrantly resistant to reducing conditions; thus, they may represent a different species compared to the biologically active disulfide-bound dimers. The relevance of an apoE dimer/monomer profile in AD was also addressed previously in plasma, with the identification of dimers only in
APOE-ɛ3 carrier subjects, the levels of which decreased in the demented group [
49]. A recent report using two-dimensional gel electrophoresis indicated that plasma apoE is elevated in AD with respect to controls [
50]. However, it is worth noting that apoE does not cross the blood-CSF barrier [
51].
ApoE can form heteromeric complexes with other apolipoproteins [
17] and with proteins such as the ciliary neurotrophic factor [
52] or APP [
53], among others, but principally with Aβ. Indeed, apoE can form in vitro SDS-stable complexes with Aβ [
1,
54,
55], but the interaction with exogenous Aβ does not induce drastic changes to the overall size of the Aβ/apoE-containing lipoprotein particles [
55]. The formation of noncovalent apoE/Aβ complexes (1:1) is implicated in both Aβ clearance and fibrillization, and the three isoforms of apoE are able to form these complexes [
56]. Complexes of apoE and Aβ have been demonstrated in non-pathological human CSF [
55] and in AD brain [
57,
58]. Thus, Aβ may act as a triggering driver for the crosslinking and stabilization of aberrant apoE complexes. In the AD brain, the balance between soluble to insoluble apoE/Aβ aggregates has been associated with impaired apoE activity in Aβ clearance, as apoE is responsible for the accumulation and fibrillization of Aβ [
59]. The effects of apoE on Aβ aggregation may be restricted to HDL-like particle-bound apoE [
60]. Other studies have demonstrated that apoE influences Aβ clearance despite minimal interaction [
61]. However, despite the fact that Aβ can contribute to the formation of stable apoE dimers as a crosslinking agent, the behavior of the resulting species may differ from other apoE/Aβ aggregates. We favor the hypothesis that the stable apoE complexes may have compromised biological activity, regardless of the presence of Aβ.
It is also interesting to note that apoE binds Aβ in an isoform-specific manner. Thus, monomeric apoE4 binds to Aβ peptide more rapidly than monomeric apoE3 or apoE2, and so it appears that the efficiency of binding correlates inversely with the risk of developing AD pathology [
62]. Moreover, soluble SDS-stable complexes of apoE4/Aβ precipitate more rapidly than apoE3/Aβ complexes [
63]. Whether these monomeric apoE/Aβ complexes trigger the formation of oligomeric complexes, and the potential compromise of the apoE peptides involved in these complexes on Aβ clearance in vivo, require analysis.
The aberrant apoE complexes may also influence the role of apoE on lipid metabolism and transport. It is assumed that unlipidated apoE monomers are the species that form disulfide-linked dimers; however, it is also believed that apoE must be properly lipidated to participate in cholesterol and lipid transport. Aberrant dimers are not linked by disulfide bonds, but we can only speculate whether these species are lipidated or not, and if the occurrence of these aberrant dimers could compromise the role of apoE regulating lipid homeostasis by mediating lipid metabolism and transport. ApoE4 is poorly lipidated compared with apoE2 and apoE3 [
64], and reduced binding affinity of apoE4 for HDL results in a greater proportion of unlipidated apoE, hence forming aggregates that can be more toxic for neurons than apoE2 and apoE3 aggregates [
65]. Since lipidation of apoE impedes aggregate formation [
66], we presume that these aberrant dimers are not lipidated; nonetheless, this possibility should be tested.
Finally, we found a correlation between the 100-kDa apoE levels and age in AD samples, which suggests that during pathological aging, apoE could be more likely to form non-disulfide-bound aggregates in the CSF. In the TgF344 rats, only the older animals showed statistically significant high apoE levels; accordingly, these AD models show an age-dependent increase of the levels of Aβ40 and Aβ42 from 6 months of age [
31].
The imbalance of apoE glycoforms and the existence of aberrant apoE aggregates in the CSF from AD individuals could be considered as a read-out of alterations of the biological activity of apoE in the brain of AD individuals. The possibility that CSF levels of apoE are under strong genetic influence by the APOE polymorphism is plausible; however, the relevance of these changes in CSF apoE levels on AD pathology remains elusive. The net increase of apoE levels in the CSF from AD individuals could be favored by aging. This increment, mainly due to the 34-kDa glycoform of apoE, which is likely hypo-sialylated, and the appearance of a β-mercaptoethanol-resistant 100-kDa apoE species, could indicate that the ability of apoE in AD to achieve its biological functions may be compromised.
In conclusion, while apoE levels tend to increase in AD CSF, this increase is more noticeable in certain glycoforms of monomers and aberrant complexes that may hinder its biological activity. A specific description of how these species affect apoE signaling and Aβ clearance should improve our understanding of the role of apoE in the AD pathology.
Declarations
Competing interests
HZ has served at scientific advisory boards and/or as a consultant for Abbvie, Alector, Annexon, Artery Therapeutics, AZTherapies, CogRx, Denali, Eisai, Nervgen, Novo Nordisk, Pinteon Therapeutics, Red Abbey Labs, Passage Bio, Roche, Samumed, Siemens Healthineers, Triplet Therapeutics, and Wave; has given lectures in symposia sponsored by Cellectricon, Fujirebio, Alzecure, Biogen, and Roche; and is a co-founder of Brain Biomarker Solutions in Gothenburg AB (BBS), which is a part of the GU Ventures Incubator Program (outside submitted work). KB has served as a consultant, at advisory boards, or at data monitoring committees for Abcam, Axon, Biogen, JOMDD/Shimadzu, Julius Clinical, Lilly, MagQu, Novartis, Prothena, Roche Diagnostics, and Siemens Healthineers and is a co-founder of Brain Biomarker Solutions in Gothenburg AB (BBS), which is a part of the GU Ventures Incubator Program, all unrelated to the work presented in this paper. JF has served as a consultant for Novartis and Lundbeck; has received honoraria for lectures from Roche, NovoNordisk, Nestle, Esteve, and Biogen; and served at advisory boards for AC Immune, Zambon, and Lundbeck. D.A. participated in advisory boards from Fujirebio-Europe and Roche Diagnostics and received speaker honoraria from Fujirebio-Europe, Roche Diagnostics, Nutricia, Krka Farmacéutica S.L., Zambon S.A.U., and Esteve Pharmaceuticals S.A. AL has served at scientific advisory boards from Fujirebio-Europe, Nutricia, Roche-Genentech, Biogen, Grifols, and Roche Diagnostics and has filed a patent application of synaptic markers in neurodegenerative diseases.