Background
Triggering receptor expressed on myeloid cells 2 (TREM2) is part of the immunoglobulin-lectin-like receptor superfamily, and in the brain, TREM2 is expressed mainly in microglial cells. TREM2 couples to DNAX-activating protein of 12 kDa (DAP12) and activates its downstream targets through the phosphorylation of the immunoreceptor tyrosine-based activation motif (ITAM) of DAP12. TREM2 function has been related to phagocytosis, cell growth, regulation of actin cytoskeleton, migration towards chemokines, and cytokine release (reviewed in [
1,
2]). The inheritance of a mutant variant of the
TREM2 gene,
R47H, confers a markedly increased risk for developing late-onset [
3,
4] and early-onset [
5] Alzheimer’s disease (AD). TREM2 expression has been found in the close vicinity of amyloid plaques in AD brain and transgenic mouse models, particularly in microglia and infiltrating macrophages surrounding plaques [
6,
7]. Moreover, loss-of-function mutations in
TREM2 are linked to an increased risk of developing Nasu-Hakola disease, frontotemporal dementia, Parkinson’s disease, and sporadic amyotrophic lateral sclerosis [
8‐
13]. This suggests a significant role for TREM2 in neurodegenerative diseases with phenotypes being dependent upon the location of the mutation and the severity of protein dysfunction [
14]. Since
TREM2 variants seem broadly involved in neurodegeneration, there is an urgent need to further investigate the functions of TREM2 in the brain and to find ligands involved in TREM2-mediated signaling and their role in AD pathogenesis.
Another strong risk factor for developing late-onset AD is the
ε4 allele of the apolipoprotein E gene (
APOE), which is far more prevalent than the
R47H-TREM2 risk factor [
3]. The three major isoforms of ApoE in human are the ApoE ε2 (cys112, cys158), ApoE ε3 (cys112, arg158), and ApoE ε4 (arg112, arg158). The most common isoform is ApoE ε3, while ApoE ε2 isoform is rare. The
APOE4 allele increases the risk of AD by three- to fourfold [
15,
16], while
APOE2 is protective [
17]. The disease risk is gene dose-dependent for both alleles [
15‐
17]. In the brain, astrocytes are the main source of ApoE, and ApoE is the major regulator of lipid metabolism, transporting cholesterol and phospholipids between cells (reviewed in [
18]). ApoE is intimately associated with fibrils of amyloid-β (Aβ) in amyloid deposits in AD brain and transgenic animal models of AD [
16,
19,
20]. Early studies reported both inhibition and augmentation of Aβ fibril formation when mixing ApoE and Aβ in vitro [
21‐
25]. Some studies reported that the ApoE ε4 isoform bound faster to Aβ and increased aggregation more readily than the ApoE ε2 or ε3 isoforms [
23,
25‐
27]. The interpretation of such studies in relation to the brain is complicated by the complex nature of Aβ (mixture of different lengths and aggregation states) and ApoE (isoform and lipidation state), which is difficult to mimic in vitro. A link between human ApoE and amyloid plaque accumulation in amyloid-β precursor protein (AβPP) transgenic mouse brain has been consistently found [
28‐
30]. Since mice have an ApoE isoform distinct from the human isoforms, these studies have mostly been done on AβPP transgenic mice crossed with
Apoe knockout mice (
Apoe
−/−) instead expressing human
APOE. Such studies have shown that murine ApoE augments Aβ deposition more than either of the human ApoE isoforms and that the order of fibrillogenic effect on Aβ of the human ApoE isoforms is ApoE ε4 > ApoE ε3 > ApoE ε2 [
28,
31,
32]. These results are consistent with the relative risk associated with the ApoE variants for developing AD. The effect of ApoE on Aβ aggregation also depends on the human ApoE concentration [
30], as also reported for murine ApoE in the first AβPP/ApoE cross-breeding study [
33].
All three main human ApoE isoforms were recently found to bind to human TREM2 in vitro [
34,
35]. In a recent paper, these findings were partly opposed, as binding could not be detected between human TREM2 and nonlipidated human ApoE using protein microarray, while lipidated human ApoE bound TREM2 when bio-layer interferometry was used [
36]. The reported experiments were all done in pure in vitro (cell-free) settings [
34‐
36]. It remains unclear whether ApoE binds to TREM2 when it resides on cell surfaces, and whether ApoE serves as a TREM2 agonist. Experimental proof of such an interaction would effectively link the two major genetic risk factors for AD in a signaling pathway.
In this study, we used a cell reporter assay to provide the first evidence of human ApoE-mediated intracellular signaling through human TREM2, which is of great importance for understanding of the interaction between ApoE and TREM2 in the AD pathogenesis. Possible interactions between human ApoE and murine TREM2 have never been examined despite multiple studies using transgenic AβPP mice expressing human ApoE instead of murine ApoE. We report that human ApoE signals through murine TREM2, albeit with a reduced efficacy. The
K
d has previously only been determined for the ApoE ε3 isoform using dot blotting [
34], a semi-quantitative method. We developed a sensitive ELISA-based binding assay to determine the affinity of human TREM2 binding to the three major human ApoE isoforms and, importantly, identified a TREM2-binding region in human ApoE, which has never been reported before.
Discussion
APOE and
TREM2 are the two most predominant genetic risk factors in AD. Previous papers have reported binding between human ApoE and human TREM2 only in pure in vitro settings [
34‐
36], but did not contain investigations on whether ApoE was an agonist mediating intracellular signaling through TREM2, or alternatively an antagonist inhibiting other stimuli. In this study, using cell reporter assays, we provide the first evidence that ApoE and TREM2 interact in a signaling pathway. With a sensitive ELISA-based binding assay, we have determined binding affinities of human ApoE ε2, ε3, and ε4 to human TREM2 to be in the low nM range. Importantly, we report the first evidence that a TREM2-binding domain is found in amino acids 130–149 of human ApoE, which is the conserved receptor-binding domain in the main ApoE isoforms, ε2, ε3, and ε4, as reviewed in [
48].
By using an ELISA-based binding assay, we found a statistically significant lower
K
d for ApoE ε4 (9.5 nM) than for ApoE ε2 and ApoE ε3 (13 and 16 nM, respectively), while the EC
50 for human TREM2-signaling was similar among the ApoE isoforms (27, 33, and 34 nM for ApoE ε2, ε3, and ε4, respectively). Since we found a TREM2-binding domain in a conserved region of ApoE (amino acids 130–149), it is reasonable to expect equal binding affinity between the recombinant human ApoE isoforms, as was also previously reported [
34,
35]. The biological relevance of potential isoform differences in TREM2 stimulation, if any, needs to be further investigated by in vivo modeling. The EC
50 data found are in the physiologically relevant range, since the ApoE concentration is between 60–300 nM in the human cerebrospinal fluid (CSF) and 880–2700 nM in the plasma [
49‐
53]. In the brain, it can generally be expected that the concentration of a given protein depends on location. The availability of a ligand to a receptor also depends on its binding to other proteins. Accordingly, ApoE is likely to be present in higher local concentrations in amyloid plaques and close to cell surfaces of astrocytes, and the CSF concentration gives only a rough estimate of the ApoE concentration in the brain. With western blots, we did not find evidence that TREM2 was overexpressed in the BWZ reporter cells as compared to the brain from a nontransgenic 14-month-old mouse. Using
Trem2 knockout brain tissue, we showed that the immunoreactive band in tissue extract of the nontransgenic mouse brain is largely derived from murine TREM2. The major limitation with the western blot experiments is that the binding of the TREM2 antibody might depend on TREM2 glycosylation, which presumably differs between native TREM2 in mouse brain and TREM2 in transfected BWZ cells, and surely with nonglycosylated recombinant TREM2. The TREM2 antibody used was chosen since it was raised against a domain (amino acids 154–165 of mouse TREM2), which is distinct from the location of the
N-glycosylation, which occurs at amino acids 20 and 80 [
54]. Yet, it still cannot be excluded that differential glycosylation still plays a role. A 40–45-kDa size of the TREM2 immunoreactive band from brain extract is quite consistent with previous observations [
55]. Glycosylation structure and complexity affects migration of proteins on a polyacrylamide gel. The vector should encode a TREM2 fusion protein of ≈36 kDa, but since it is expressed in a cell line, the protein glycosylation is likely to be less complex than that of the native TREM2 protein in mouse brain.
By stimulating BWZ-hTREM2 reporter cells with ApoE, we found a significantly higher Hill coefficient for ApoE ε4 (1.7) than for ApoE ε2 (1.0). When using a sensitive ELISA-based binding assay, we found similar Hill coefficients between the ApoE isoforms when bound to recombinant, human TREM2. The relevance of binding differences needs to be further explored with other experimental techniques such as structural modeling of human TREM2 bound by ApoE isoforms.
It might be tempting to conclude that the risk of AD does not depend on ApoE/TREM2 signaling since we did not find evidence of ApoE isoform-dependent differences in binding to or signaling through TREM2. However, other factors could play a role in vivo: (1) It has been shown that the ApoE concentration in CSF depends on
APOE genotype [
51], making it plausible that individuals homozygous for
APOE4 exhibit diminished stimulation of TREM2 as a result of lower ApoE concentration; (2) the outcome of the ApoE-mediated intracellular signaling through TREM2 could vary depending on ApoE isoforms due to cofactors with different binding affinities exerting effects on structure and stability, or perhaps even downstream targets of the intracellular signaling through TREM2. Thus, it is still possible that the ApoE/TREM2 interaction contributes to the ApoE isoform-dependent AD risk.
Ligand-induced receptor signaling also depends on interactions with other constituents in the membrane and on anchoring proteins. Heparan sulfate proteoglycan (HSPG) is a possible cofactor that could be involved in the ApoE stimulation of TREM2. HSPGs are ubiquitously present on cell surfaces and in extracellular matrices, and the heparan sulfate (HS) side chains can bind ApoE [
56‐
58]. Interestingly, both the receptor-binding region and the lipid-binding region of ApoE are involved in binding to HS [
57‐
63]. HSPGs are known to regulate the interaction between ligands and their receptors [
64‐
67]. In this sense, ApoE ε2 binds more readily to HS than ApoE ε3 [
56,
68], thus perhaps differentiating the ApoE ε2-mediated activation of TREM2 from the ApoE ε3-mediated activation. Interestingly, HS and ApoE are both present in amyloid plaques in AD and animal models [
69‐
73].
Another possible cofactor for the ApoE/TREM2 signaling is Aβ. While Aβ has been reported to bind both the lipid-binding region (amino acids 244–272) and the receptor-binding region (within amino acids 130–149) of ApoE [
26,
74], it is interesting to note that Aβ binding to the lipid-binding region of ApoE would allow for the possibility of stimulation of TREM2 by the Aβ/ApoE complexes found in or around amyloid plaques.
A link between ApoE and Aβ has been found consistently, as
APOE dose-dependently enhances the risk of developing late-onset AD and also increases amyloid burden among those carrying
ε4 alleles [
15,
75]. Likewise, in APP
V717F transgenic mice, the ApoE ε4 isoform was found to increase Aβ deposition [
28] and decrease Aβ clearance [
32], although the specific clearance pathway involved is still incompletely understood. In C57BL/6 mice, ApoE (lipid-poor or lipid-rich) and Aβ (monomeric Aβ
40 or Aβ
42) were injected intracerebrally either alone or in ApoE-Aβ complexes. The authors found decreased Aβ clearance associated with the ApoE ε4 isoform [
76]. Since ApoE is associated with Aβ in amyloid plaques and TREM2 is found in microglia or macrophages adjacent to the plaques [
6,
7], it is intriguing to think that there might be a connection between Aβ/ApoE complexes and TREM2 on surrounding microglia/macrophages.
While reduced ApoE gene dosage consistently ameliorates AD phenotypes [
30,
33], there has been conflicting reports on gene dosage effects of TREM2 in AβPP transgenic mice [
7,
77,
78]. Reduced TREM2 gene dosage was found to enhance amyloid pathology in one of the studies [
78], while it decreased amyloid pathology in another study [
7]. Importantly, the latter group recently showed that TREM2 deficiency in APP/PS1 mice lead to reduced amyloid burden at an early disease stage, while it enhanced amyloid burden at a late disease stage [
79]. When AD pathogenesis is investigated in transgenic mouse models, human ApoE isoforms are often expressed, either by using a heterologous promoter [
80] or by gene replacement in animals devoid of murine ApoE [
81]. Knowing that the TREM2/ApoE interaction is presumably relevant to AD pathogenesis and perhaps also to other neurodegenerative diseases, we found it essential to determine the extent to which human ApoE stimulated murine TREM2 signaling. Such information is needed to better interpret studies with mouse models of disease. We found that all three human ApoE isoforms were agonists to murine TREM2 albeit with slightly reduced efficacy compared to human TREM2. Therefore, double transgenic ApoE/AβPP animal models might inadequately reflect the interactions between ApoE and TREM2 in the AD brain; microglial activation triggered by ApoE stimulation of murine TREM2 might be more modest in transgenic mice than when TREM2 in AD brain is exposed to the same stimuli. Thus, it could be worth the effort to create human
TREM2 knock-in models, or at least to insert exons encoding the extracellular domains of human
TREM2 into the murine genome.
The ApoE/TREM2 interaction as observed by us might be involved in a clearance pathway, which is consistent with experiments with transgenic mice convincingly linking ApoE to Aβ clearance [
28,
32]. ApoE could potentially also regulate responses of macrophages or microglia to specific stimuli of neuroinflammation and lipid metabolism and activate specific cellular functions such as phagocytosis. Microglia do not express TREM2 constitutively on the cell surface [
82], thus complicating investigations of ligand-receptor interactions with primary cells in vitro. The reporter cells used here serve as an efficient and sensitive tool delivering reliable quantitative data on ligand-receptor interactions and signaling. However, there might be differences in the cell-surface expression of proteins or glycoproteins (possibly functioning as cofactors to TREM2) as well as intracellular signaling cascade between reporter cells and microglia/macrophages.
In the plasma, the ApoE isoforms differentially bind to lipids, with ApoE ε2 and ε3 preferring high-density lipoprotein (HDL) lipids and ApoE ε4 preferring low-density lipoproteins (LDL) and very low-density lipoprotein (VLDL) lipids [
48,
83‐
86]. However, the intact BBB of the adult brain confers a barrier through which neither lipoproteins nor lipids can pass, thus creating a restrictive environment that is distinct from the plasma. In this manner, LDL and VLDL are not present in the CNS, and the astroglial-derived HDL particles are structurally different from those found in the plasma (see [
18] for a review on CNS lipoproteins). However, the lipoprotein-affecting enzymes, transporters, and receptors present in the periphery are also found in the CNS [
18], and different binding affinities have been reported for the ApoE isoforms to their receptors, which can also be affected by ApoE lipidation state for some receptors [
76,
87,
88]. Using reporter cells, it was recently found that various anionic lipids found in AD amyloid plaques were agonists for TREM2 [
78]. Since ApoE is a lipid carrier, the ApoE/TREM2 interaction is likely to be involved in lipid sensing by TREM2 [
78,
89], although we hereby demonstrate that the nonlipidated ApoE protein is sufficient for a high-affinity interaction. Initial studies of TREM2 ligand binding suggested that TREM2 could function as a pattern recognition receptor distinguishing negatively charged motifs [
37]. In addition to ApoE binding, other apolipoproteins, ApoA1, ApoA2, ApoB, and ApoJ, also bind to TREM2 [
34‐
36]. Binding of lipidated as well as nonlipidated human ApoE to human TREM2 has been demonstrated [
34‐
36]. In our cell reporter assays, ApoE is bound to a Maxisorp plate, and it is unlikely that lipids from cells or FBS will lipidate ApoE under these circumstances. Also arguing against the necessity of ApoE being lipidated in order to bind TREM2 is our demonstration of high affinity binding between recombinant ApoE and recombinant TREM2 in a cell-free and lipid-free ELISA binding assay. Moreover, the dose-response curves for in vitro binding data and reporter cell signaling data were quite similar, which further argues against lipidation of ApoE being necessary for TREM2 signaling. Finally, the ApoE fragment, ApoE-mim149, binds to TREM2. This peptide is derived from the receptor-binding domain of ApoE, which is distinct from the lipid-binding domain [
48]. It is conceivable that immobilization of ApoE to a plate can affect its conformation such that an ApoE conformer is created, which is prone to stimulate TREM2 in the reporter assay. However, ApoE interacting with TREM2 and displaying EC
50 and
K
d values in the same approximate concentration range when immobilized (reporter assay) and when in solution (binding and competition assays) argues that such a potential effect is either very limited or not relevant in our experiments.