High affinity interactions and signal transduction between Aβ oligomers and TREM2

Rare coding variant in the Triggering receptor expressed on myeloid cells 2 (TREM2) are associated with increased risk for Alzheimer’s disease (AD), but how they confer this risk remains uncertain. We assessed binding of TREM2, AD associated TREM2 variants to various forms of Aβ and APOE in multiple assays. TREM2 interacts directly with various forms of Aβ, with highest affinity interactions observed between TREM2 and soluble Aβ42 oligomers. We confirm the previous interaction between APOE3 and APOE4 and TREM2. High affinity binding of TREM2 to Aβ oligomers is characterized by very slow dissociation. Pre-incubation with Aβ is shown to block the interaction of APOE. In cellular assays, AD-associated variants of TREM2 reduced the amount of Aβ42 internalized, and in NFAT assay the R47H variant decreased NFAT signaling activity in response to Aβ42. These studies demonstrate i) a high affinity interaction between TREM2 and Aβ oligomers that can block interaction with another ligand and ii) that AD-associated TREM2 variants bind Aβ with equivalent affinity but show loss of function in terms of signaling and Aβ internalization.


INTRODUCTION
The most prevalent form of dementia, Alzheimer's disease (AD), is hypothesized to be triggered by accumulation of aggregated amyloid-β (Aβ) followed by a "cascade-like" chain of events that includes induction of tauopathy, neurodegeneration and alterations in innate immune signaling (Hardy & Selkoe, 2002;Musiek & Holtzman, 2015). The later feature is reflected by the presence of a reactive astrocytosis, microgliosis and increased level and accumulation of a variety of immune molecules. A more central role of the innate immune system and microglial cells in particular, has emerged from genetic studies. These studies show that numerous loci encoding immune genes are associated with altered risk for AD. Even more compelling are data showing that coding variants in three different genes (TREM2, PLCG2, and ABI3), whose transcripts are expressed primarily in microglial cells in the brain, alter risk for AD (Golde et al, 2013;Jin et al, 2014;Sims et al, 2017). Despite this risk, there is little consensus as to how the alterations in the immune system mechanistically alter AD risk. Modeling studies reveal a complex relationship between immune activation states, the proteinopathies found in AD, and neurodegeneration -a phenomenon we refer to as immunoproteostasis (Chakrabarty et al, 2015). Indeed, at present, there is still no definitive insight into how genetic variation within immune loci, alters function of the immune system, and thereby alter risk for AD. Further understanding of how genetic risk factors impact the immune function will be important to guide therapeutic development, that to date has largely focused on anti-inflammatory strategies (Chakrabarty et al, 2015).
The rare protein coding variants in TREM2, R47H and R62H, which are reproducibly associated with increased risk for developing AD, are being intensively studied as these variants were the first coding variants in an innate immune gene that have been reproducibly shown to alter risk for AD (Cruchaga et al, 2013;Jin et al, 2015;Lill et al, 2015). Prior to association with AD, homozygous or compound heterozygous mutations in TREM2 had been identified in a disease called Polycystic Lipomembranous Osteodysplasia with Sclerosing Leukoencephalopathy (PLOSL) or Nasu Hakola disease (Bianchin et al, 2006). PLOSL is characterized by fractures, frontal lobe syndrome, and progressive presenile dementia beginning in the fourth decade. Numerous studies of the PLOSL-associated TREM2 variants suggest that these variants are loss of function altering maturation and cell surface expression of TREM2 (Paloneva et al, 2001;Paloneva et al, 2002). Both structural and cell biology studies of the AD TREM2 variants show they lie on the surface of the TREM2 ectodomain, possibly framing a binding pocket, whereas the PLOSL variants are within the core structure (Kleinberger et al, 2014;Kober et al, 2016). Maturation studies demonstrate folding and maturation deficits induced by PLOSL associated TREM2 variants, whereas the AD variants do not consistently show these same functional impairments. Notably, a DNA polymorphism within a DNase hypersensitive site 5' of TREM2, rs9357347-C, also associates with reduced AD risk and increased TREML1 and TREM2 levels (Carrasquillo et al, 2017). This variant is associated with decreased risk for AD, which would be consistent with the hypothesis that the coding variants in TREM2 associated with increased AD risk are partial loss of function.
TREM2 is a type I transmembrane glycoprotein that interacts with the transmembrane region of DAP12 (TYROBP) to mediate signaling events through DAP12's immunoreceptor tyrosine-based activation motif (ITAM) domain (Peng et al, 2010). The extracellular, immunoglobulin-like domain V type, domain of TREM2 is glycosylated and shows cell surface localization. TREM2 is highly expressed in microglia and peripheral macrophages . TREM2 is one member of a family of receptors that has diverged between mouse and humans. In both mouse and human the family members are clustered within a single chromosomal region on chromosome 6. In the context of neurodegenerative disease, the other family members have not been intensively studied as they are typically expressed at much lower levels in the brain. Like many type I membrane receptors, TREM2 can undergo ectodomain shedding and subsequent intramembrane cleavage by -secretase (Wunderlich et al, 2013). The physiologic implications of these proteolytic cleavages remain unclear, though increased soluble TREM2 has been reported in the CSF of early stage Alzheimer's disease (Suarez-Calvet et al, 2016), and study of a variant p.H157Y reported to be associated with AD in the Han Chinese population has been shown to increase ectodomain shedding (Jiang et al, 2016;Schlepckow et al, 2017;Thornton et al, 2017). Enhanced shedding activity from the mutation p.H157Y located in the ectodomain cleavage site leads to a reduced cell surface TREM2 and phagocytic activity (Schlepckow et al, 2017;Thornton et al, 2017).
TREM2 has been implicated in multiple functions including migration, survival and proliferation, cytokine release, and phagocytosis (Kleinberger et al, 2014;Yuan et al, 2016). TREM2 appears to be a promiscuous receptor and has been shown to interact with cellular debris, bacteria, anionic phospholipids, nucleic acids and Apolipoprotein E (APOE) (Atagi et al, 2015;Bailey et al, 2015;Kober & Brett, 2017;Song et al, 2017;Yeh et al, 2016). Alterations in signaling and binding of TREM2 AD variants compared to wt TREM2 have been reported, though the pathogenic relevance of these alterations is not yet completely clear (Kleinberger et al, 2014;Kober et al, 2016). Mouse modeling studies show increased TREM2 expression in multiple models of neurodegenerative diseases, and TREM2 positive microglial cells appear to surround plaques in APP models (Jay et al, 2017;Jiang et al, 2014;Ulrich et al, 2014;Yuan et al, 2016). TREM2 also shows an altered distribution in microglia surrounding the amyloid plaque, with much of the TREM2 immunoreactivity localized within the plasma membrane of the microglial cells adjacent to the plaque. In TREM2 knockout mice crossed with APP there are fewer microglial cells surrounding plaques and an increase in plaques with "specular" amyloid morphology as well as more neuritic dystrophy Wang et al, 2016). Such data has led to the hypothesis that TREM2 may regulate the interaction of microglia with plaques and protect other brain cells from toxicity.
In this study, we investigated if TREM2 and its AD variants interact directly with Aβ. We found no significant changes in the binding affinity between Aβ42 with TREM2 or the AD variants or with their mouse counterpart Trem2, Trem1 and Treml1. Despite this lack of difference in the binding affinity, we observed an attenuated activation of NFAT signaling and a reduced amount of Aβ42 internalized by cells expressing the TREM2 AD variants. We show that TREM2, TREM1 and TREML1 bind APOE and Aβ variably. Our study reveals that the TREM2 and other family members can bind various forms of Aβ and that Aβ oligomers can induce singling events. Further, our data are consistent with the hypothesis that AD-risk variants in TREM2 induce a partial loss of function (Song et al, 2018), but do not directly alter binding affinity for either Aβ or APOE.

Soluble TREM2-Fc and soluble TREML1-Fc interact with Aβ42 fibrils.
The ectodomain of TREM2 interacts with various ligands including bacterial liposaccharides and phospholipids. We explored whether TREM2 might interact with A and whether this interaction might be affected by the AD variants. We also analyzed mouse Trem2 in order to compare its relative binding ability to human TREM2. Initial studies focused on whether TREM2 could interact with fibrillar Aβ42 (fAβ), using an Aβ42 pulldown assay we have previously validated (Chakrabarty et al, 2015). Conditioned media from transiently transfected HEK 293 cells expressing the soluble ectodomains of i) human TREM2 fused to the Fc domain of human IgG4 (sTREM2-Fc), ii) AD associated TREM2 variants (sTREM2-R47H-Fc, sTREM2-R62H) or iii) mouse TREM2 (sTrem2-FC) was incubated with fAβ and spun at 18000 X g to pellet fAβ and associated proteins, sTREM2 and sTREM2-R47H and sTREM2-R62H were enriched in the fAβ pellet ( Fig 1A), whereas the Fc control was not detected. Mouse sTrem2 also showed binding to A fibrils. Similar pulldown assay between fAβ and full length TREM2 were conducted using the cleared RIPA lysate from TREM2, TREM2-R47H and TREM2-R62H transiently transfected into HEK293T cells.
Although less efficient, presumably because of detergent present in the lysate, full length TREM2 and the AD variants were pulled down by fAβ ( Fig 1B).
Multiple studies show TREM2 ectodomain interacts with APOE (Atagi et al, 2015;Bailey et al, 2015;Yeh et al, 2016). To confirm that the sTREM2-Fc proteins are functional and bind APOE3, we co-transfected TREM2-Fc constructs with APOE3 and pulled down proteins that bound sTREM2-Fc with anti-IgG Fc agarose beads. These data show that the TREM2 ectodomain binds APOE and that there is no major difference in binding between the AD variants and wt TREM2 (Fig 1C). In this experiment, a minor interaction between the human sIgG4-Fc domain and APOE3 is observed.
TREM2 is a member of a multiprotein family that is divergent between human and mice. We explored whether select members of the TREM2 family bind Aβ42 fibrils and APOE. We repeated the fAβ pulldown experiments in Fig 1A with two other soluble family members, which share low sequence homology with each and other: TREM1-Fc and TREML1-Fc. In Fig 1D, we confirm presence of soluble TREM2-Fc and TREML1-Fc in the pelleted Aβ42 fibrils. Soluble TREM1-Fc was not detected in the pelleted fAβ. Soluble TREM1-Fc like soluble TREM2-Fc interacts with APOE3, but there is lack of detectable interaction between soluble TREML1-Fc and APOE3 ( Fig 1E).

Fc family members interact with Aβ42 oligomers with a different affinity.
We used BioLayer Interferometry (BLI) to further explore the interaction between A and TREM2 and other TREM family members. By immobilizing the TREM proteins on the BLI sensor, ligand interaction can be monitored in real time, enabling precise determination of the association and dissociation constants. As poorly soluble fAβ does not produce detectable binding in this assay even with high affinity control anti-A antibodies, we were only able to assess binding with Aβ42 oligomers, and A40 and Aβ42 monomers. We initially explored interactions between Aβ42 oligomers and sTREM2-Fc in the BLI assay. These data revealed that sTREM2 bound Aβ42 oligomers with high affinity largely attributable to a very slow dissociation. In these initial studies, we also found no interaction between the Fc alone and Aβ42 oligomers. Further, when a high affinity anti-Aβ1-16 antibody (Ab5) was incubated following the dissociation step, we were able to demonstrate that the Aβ oligomers were indeed tightly complexed with sTREM2 on the sensor (Supplemental figure 2A). After further optimization of the assay (see methods), we systematically evaluated the binding of sTREM-Fc proteins to various concentrations of Aβ42 oligomers, Aβ42 monomers Aβ40 monomers, as well as lipidated recombinant APOE3 and APOE4 (Fig 2A and B, Supplemental figure 2b). Under these optimized conditions, no sFc binding is observed, and non-specific binding of APOE3 or 4 eliminated. These studies enabled precise determination of the Kon, Kdis (Koff) and the overall KD ( Fig 2C) of various form of Aβ and APOE under identical conditions. These studies revealed very high affinity binding between Aβ42 oligomers and sTREM2-Fc and the AD variants (<1.0E-12 M). This strong binding is largely attributable to the apparent irreversibility of the binding. Indeed, the dissociation constants are the maximal measurable (kdis = 1.0E-07 1/s) on the Octet Red instrument. In contrast to the binding observed between sTREM2 and Aβ42 oligomers, binding to Aβ42 monomers was much weaker and reversible (KD from 1.4 to 2.9E-07 M). The KDs of the wt TREM2 were not statistically different from the AD variants. Binding to Aβ40 monomers was similar to binding to Aβ42 monomers (KDs from 1.4 to 2.7E-07 M), and again the KDs of the wt TREM2 were not statistically different from the AD variants. We also established KDs in the nanomalor range for binding of lipidated APOE3 and 4 with sTREM2. KDs for APOE3 binding sTREM2 or the AD variants were very similar and not statistically different. Binding of APOE4 revealed a slight increase in binding affinity of R47H variant (1.7 fold) and the R62H variant (1.56 fold) with APOE4 that was statistically significant. Given the irreversible nature of the Aβ42 oligomers binding to sTREM2, we explored whether the binding of Aβ42 oligomers could block interaction with APOE3. We evaluated whether pre-incubation with Aβ42 oligomers to sTREM2 could block subsequent interaction with APOE3. As shown in Fig 3, binding of Aβ42 oligomers blocked APOE3 binding. In contrast addition of Aβ42 oligomers to sTREM2 previously bound to APOE3 did not block subsequent interaction with Aβ42 oligomers. We next examined the binding of select soluble mouse and human TREM family members with Aβ42 oligomers by BLI ( Fig 2D). TREML1 showed a similar high affinity binding of Aβ42 oligomers to TREM2 but the interaction between TREM1 and Aβ42 oligomers was much weaker (1.1E-7 M). In contrast, both mouse Trem1 and Treml1 bound Aβ42 oligomers with high affinity.
We further investigated the binding of fAβ with soluble TREM2-Fc and the AD variants by an ELISA binding assay due to their unsuitability with BioLayer Interferometry. We compared the binding of fAβ, Aβ42 monomers and Aβ42 oligomers with soluble TREM2-Fc and the AD variants. In Supplemental figure   3, Aβ42 fibrils incubated with soluble TREM2-Fc and the AD variants or soluble Trem2-Fc increased weakly the signal compared to the Fc control and showed no significant differences. The Aβ42 oligomers incubated with soluble TREM2-Fc or the AD variants or soluble Trem2-FC produced a signal significantly higher compared to Fc control and around two times stronger than the one from the Aβ42 monomers.
In Fig 3A, we tested Aβ42 monomers at higher concentrations. A significant signal between Fc control and the soluble TREM2-Fc or the AD variants or soluble Trem2-Fc is obtained at 1000 nM Aβ42 monomers and more. This result confirms the binding of Aβ42 monomers with soluble TREM2 and the AD variants and soluble Trem2-Fc without significant differences between all of them. We then repeated a similar ELISA assay with soluble TREM-Fc family members and Aβ42 oligomers. In Fig 3B,  Collectively, these biochemical experiments confirm a direct interaction between Aβ42 and TREM2 and the AD variants TREM2 R47H or TREM2 R62H and Trem2 without any significant differences. Aβ42 oligomers have a stronger affinity for TREM2 than Aβ42 monomers. TREML1 also binds Aβ42 oligomers, but failed to bind APOE3 contrary to TREM2 or TREM1.

Expression of TREM2 activates NFAT signaling and increases internalized Aβ42.
We investigated whether the binding of Aβ42 oligomers with TREM2 activates cell signaling by exposing 2B4 NFAT-GFP reporter cells transduced with TREM2 and exposed to various concentrations of Aβ42 oligomers. Control cells bound Aβ42 oligomers and weakly activated the GFP reporter compared to TREM2 expressing cells (Fig 5A). TREM2 R47H significantly activated the GFP reporter at the highest concentration of Aβ42 oligomers, but at each concentration the signal was much lower than the signal in wt TREM2 transduced cells. These data demonstrate that TREM2 activates NFAT signaling in presence Aβ42 oligomers, but this activation is strongly reduced by the AD variant TREM2 R47H.
TREM2 expression in non-phagocytic cells confers phagocytic activity (Hsieh et al, 2009). To evaluate if TREM2 expression induces Aβ42 uptake, we incubated Aβ42 over HEK293T cells transiently transfected with TREM2 and quantified internalized Aβ42. Compared to control transfection, TREM2 expression increased the amount of internalized Aβ42. TREM2 R47H and R62H also increased internalization levels relative to control, but the amount internalized was much less than that observed with wt TREM2.
We conducted several studies to further explore the reduced uptake observed by the AD variants. First, we investigated if the shedding of TREM2 is affected by Aβ. If Aβ binding induces shedding of the TREM2 ectodomain and the AD variants are shed more, this could explain the decreased uptake observed.
However, as observed in Fig 4C, despite similar levels of membrane associated TREM2, TREM2 R47H and TREM2 R62H, the amount of soluble TREM2 was similar and not altered by oligomeric Aβ. We also rigorously evaluated cell surface levels of TREM2 using cell surface biotinylation studies. These studies ( Fig 4D), again revealed no significant differences in levels as previously observed (Kleinberger et al, 2014;Kober et al, 2016).

DISCUSSION
We find that TREM2 binds to various forms of Aβ with a very high-affinity interaction with Aβ42 oligomers. Further, TREM2 appears capable of mediating Aβ internalization and Aβ oligomers induce NFAT singling through TREM2. These data establish a direct link between Aβ aggregates, which are implicated as initiators of the pathological cascade in AD and TREM2 a microglial immune receptor. No difference in binding affinity between the AD associated variants and wild-type TREM2 to any form of Aβ is detected. However, AD associated TREM2 variants appear to reduce both Aβ internalization and downstream NFAT signaling. Consistent with other published data, these data indicate that TREM2 variants associated with AD risk may act through subtle loss of function (Song et al, 2018). This data is also consistent with the reported protection from AD attributable to a non-coding nucleotide polymorphism in the TREM locus that increases both TREM2 and TREMl1 expression (Carrasquillo et al, 2017). Cell-surface levels of the AD-variants and wild type TREM2 are similar, and there is no detectable difference in internalization or processing either in the presence or absence of Aβ. Future studies will be needed to better understand the functional differences between the AD-variants and wild type TREM2, which our data strongly suggest are downstream of the initial binding event.
The differential binding affinities between various forms of Aβ and TREM2 are interesting. Binding to monomeric Aβ is relatively weak, whereas binding to oligomers is essentially irreversible. Indeed, prior interaction of Aβ oligomers blocks binding of a second ligand, lipidated APOE. We have been unable to directly assess affinities to fAβ by BLI due to technical reasons, but do observe strong interaction in the fAβ pulldown assay in the absence of detergent. In the presence of detergent or by ELISA, which also contains small amount of detergent in the binding buffer, we observe a weaker interaction. Collectively these data support a model in which various forms of Aβ interact with TREM2 with different affinity and that difference in affinity appears to be attributable to conformational differences.
Given i) the polarized localization of TREM2 on the microglial cells surrounding amyloid plaques (Jay et al, 2017;Jiang et al, 2014;Yuan et al, 2016) and ii) the lack of homing of microglial to plaques that is observed in knockout mouse models (Jay et al, 2017;Jay et al, 2015;Ulrich et al, 2014;Wang et al, 2015;Wang et al, 2016;Yuan et al, 2016), it is intriguing to consider that high affinity interaction between Aβ aggregates and TREM2 could underlie these observations. Although Aβ oligomers are described as soluble, immunohistochemical studies with select anti-oligomeric antibodies suggest that Aβ in an oligomeric conformation is present within plaques (Koffie et al, 2009). This raises an interesting speculation that TREM2 could interact with this plaque associated oligomeric conformer of Aβ resulting in cell-surface polarization and microglial homing (Tanzi, 2015;Yeh et al, 2017). Given the promiscuous ligand specificity of TREM, and the apparent ability of Aβ oligomers to block subsequent ligand binding, it is hard to predict what the biologic effects of engagement of TREM2 by Aβ oligomers would be in the peri-plaque region. Depending on relative concentration of ligands and relative efficiency of ligand induced signaling the net result could be either prolonged TREM2 activation or reduced TREM2 activation.
Human TREM2 and mouse Trem2 are 68% identical and 76% homologous, and they bind Aβ oligomers with similar affinity. This data would suggest that genetic manipulations of mouse Trem2 are likely to be informative with respect to the interactions with Aβ. Although TREM2 is the most abundant TREM family member based on expression levels in the brain, TREML1 is present at appreciable levels (Carrasquillo et al, 2017). Further, mouse Treml1 expression in the brain is dramatically increased by the transgene inserted to knockout TREM2 in one of the TREM2 mouse knockout models (Kang et al, 2018).
Given that Treml1 shows strong interaction with Aβ oligomers, this could contribute to the discrepancies reported in the various knockout models with respect to effects on amyloid deposition. Our studies of human TREM1 and TREML1, show that TREML1 binds Aβ oligomers with high affinity but TREM1 does not. Conversely, TREM1 binds lipidated APOE but TREML1 does not. In contrast both mouse Treml1 and Trem1 bind Aβ oligomers. These data can provide a basis for understanding both the larger biological role of TREM2 and family members in AD but also provide a framework to understand the structural features of the various TREM molecules required for Aβ interaction.
We confirmed the previously reported binding between TREM2 and APOE, although some of the details that we report regarding binding are distinct (Atagi et al, 2015;Bailey et al, 2015;Yeh et al, 2016).
Whereas other groups have reported reduced binding of APOE with TREM2 AD variants (Atagi et al, 2015;Bailey et al, 2015;Yeh et al, 2016), we find little difference between the binding affinity between lipidated APOE and TREM2 or the AD variants. In fact, for APOE4 binding, we see slightly increased in affinity for the AD variants compared to wild type TREM2. Though statistically significant, this is relatively small difference and unlikely to translate into biological differences with respect to function.
Differences between our data and previous reports may be due to technical differences. Yeh et al. make claims about binding affinity based on BLI but did not actually measure association and dissociation constants (Yeh et al, 2016). Yuka et al., evaluated binding of TREM2/APOE by dotblot and ELISA assays, which again do not provide precise data on kinetics of binding, and Bailey et al. reported that TREM2 R47H reduced the affinity of TREM2 for APOE used ELISA assay (Atagi et al, 2015;Bailey et al, 2015). In contrast to Yuka et al., who reported a lack of interaction between TREM1 and APOE in dot blot assay, we detect an interaction between secreted TREM1 and APOE3 from media. (Atagi et al, 2015).
Reports exploring TREM2 deficiency in mouse models of Aβ deposition have shown both differing effects on Aβ levels during early and late stages and in some cases no change in overall amyloid loads (Jay et al, 2017;Jay et al, 2015;Ulrich et al, 2014;Wang et al, 2015;Wang et al, 2016). Studies that showed effects on Aβ deposition, reported changes in Aβ loads were rather modest and regional in nature (Ulrich et al, 2014;Wang et al, 2016). More convincing are the data showing alterations in plaque morphology, specifically an increase in plaques with a spicular amyloid morphology and increased number of plaque-associated in dystrophic neurites (Jay et al, 2017;Jay et al, 2015;Jiang et al, 2014;Wang et al, 2015;Yuan et al, 2016). Given this modeling data, we would suggest that any interaction between TREM2 and Aβ (or perhaps even co-deposited APOE) in the peri-plaque region alters microglial response to the plaque but would have minimal effect on plaque clearance. Notably, recent data from one of our group shows that R47H TREM2 variant fails to restore microglial homing and activation around amyloid plaques in the mouse knockout background (Song et al, 2018). They also found that sTREM2-R47H despite being present at equal levels as a common variant (CV) of sTREM2, was not associated with plaques or neurons, whereas the sTREM2-CV was. This data presents a conundrum given that we showed kinetics of binding of the R47H and wild-type TREM2 to Aβ are indistinguishable.
Further, they are hard to reconcile functionally with the increased shedding of the han Chinese TREM2 variant that is reported to increase ectodomain shedding, but would not be expected to alter the functionality of the ectodomain (Schlepckow et al, 2017;Thornton et al, 2017).
We present biochemical and cellular data that demonstrates that TREM2 and select family members can bind to various forms of Aβ with the highest affinity interaction observed for Aβ oligomers, and we confirm an interaction between TREM2 and APOE (Atagi et al, 2015;Bailey et al, 2015;Yeh et al, 2016).
Intracellular signaling elicited following binding by Aβ oligomers is attenuated by AD associated TREM2 variants, though the AD variants appear to bind both Aβ and lipidated APOE with similar affinity to interaction with wild type TREM2. These data link TREM2 to physical interactions with two key molecules in Alzheimer's disease, and provide clues that can illuminate the clearly complex biology of TREM2 in health and disease.

Antibodies
were centrifuged at 18000g for 5 minutes at 4 degrees to remove insoluble materials. Supernatants were incubated with Aβ42 fibrils for one hour at room temperature. Aβ42 fibrils were centrifuged at 18000g for 5 minutes at 4 degrees to pellet the fibrils. Fibrils were washed with RIPA buffer or PBS (for conditioned media samples) and centrifuged again. The pelleted Aβ42 fibrils were then dissolved in protein loading buffer (31.25 mM Tris-HCl pH7.5, 2% LDS, 10% glycerol, 1.5% β-mercaptoethanol and Orange G). Cell conditioned media samples or cell lysates were also incubated with the fibril assay buffer to verify proteins aggregations. Dissolved fibrils were heated at 95 degrees for 5 min and loaded on Bis-Tris precast gels (BioRad, CA, USA) and transferred on PVDF membrane for Western Blotting. TREM2 was detected with HA antibody, soluble trems-Fc-V5 were detected with V5 antibody and fibrils detected with 6E10 antibody.

Western Blotting.
PVDF membranes were blocking in TBS 0.5% casein a one hour at room temperature. Primary antibody diluted in TBS with 0.2% Tween-20 (TBS-T) was incubated on the membrane one hour at room temperature. The membranes were washed 3 times 5 minutes with TBS-T. Secondary antibodies were diluted in TBS-T and incubated one hour at room temperature. The membranes were washing 3 times 5 minutes with TBS-T and then analyzed with Odyssey infrared imaging system (LI-COR Inc., NE, USA).

Purification of secreted soluble TREM-Fc-V5 family members from cell conditioned media.
Conditioned media from HEK293T transfected cells was centrifuged 5 min at 800g to clear cells and then transferred to a tube with 0.5% Triton X100, complete protease inhibitors, and 5 mM EDTA. Samples were incubated 2 hours at room temperature with Anti-Human IgG Fc Specific agarose beads (Sigma, MO, USA) and washed three times with PBS. Purified proteins from agarose beads were washed in PBS and eluted with 100 mM glycine pH 2.7 followed a neutralization (pH~7.5) with 1M Tris-HCl pH 8.8.
Alternatively, agarose beads were pre-blocked with 5% Bovine Serum Albumin (Sigma, MO, USA) 2 hours at room temperature and washed with PBS before the purification. Purified samples were verified by Western blot to confirm protein purification.

APOE interaction assay.
HEK293T cells were co-transfected with soluble TREM2-Fc family members and APOE3. The soluble TREM-Fc family members were purified as described above with BSA pre-blocked agarose beads Anti-Human IgG Fc Specific. Agarose beads were washed with PBS 0.5% Triton and eluted with protein loading buffer. Eluates were loaded on Bis-Tris precast gels and transferred on PVDF membrane for Western Blotting. Soluble TREM2 Fc-V5 and AD variants were detected with V5 antibody, and APOE3 detected with Ab947 antibody.

BioLayer Interferometry.
Soluble TREM2-Fc-V5 and TREM2-Fc-V5 family proteins were purified as describe above with antihuman IgG Fc Specific agarose beads. Aβ42 monomers and oligomers were prepared as described (Stine et al, 2003).Kinetic determination of TREM2/Aβ42 oligomer was initially performed on the Octet

ELISA binding assay.
Soluble TREM-Fc family member proteins were purified and eluted as previously described for the BioLayer Interferometry. Purified samples were coated overnight at 4 degrees in a 96 well plates (Immulon 4HBX, Thermo Scientific, NY, USA) in 100mM NaCO3 pH 11. Plates washed with PBS twice and then blocked three hours at room temperature with blocking buffer (Block Ace, Abd Serotec, Japan).
Blocked samples were incubated with Aβ42 monomers, Aβ42 oligomers or Aβ42 fibrils at specified concentration for one hour at room temperature. Aβ was detected with 4G8-HRP and soluble TREM-Fc samples were detected with anti-human IgG HRP to insure similar coating level of proteins. Samples were developed with TMB substrate (KPL, Maryland, USA) and the reaction was stopped by adding 6.67% O-phosphoric acid (85%). The signal produced by HRP and TMB substrate was quantified by measuring the absorbance at 450 nM.

Aβ cellular uptake.
HEK293T transfected cells from a 6 well plate were incubated 3 hours with 500 nM Aβ42 monomers.
Cells were washed two times with PBS, incubated with trypsin 5 min and centrifuged 3 min at 500g.
Supernatant was removed and cells were washed three times with a large volume of PBS (10-15 ml/wash). Pelleted cells were lysed with 100 ul SDS 2% for 5 min and then 200 ul of PBS was added.

Biotinylation.
Biotinylation experiments were performed to detect cell surface expression of TREM2 and the AD variants. Briefly, transfected HEK293T cells were incubated with 2 mg/ml sulfo-NHS-biotin (Pierce) in PBS for 30 minutes on ice. Cells were washed with PBS 100 mM glycine, PBS and then lysed in RIPA buffer. Cell lysates were incubated with streptavidin agarose resin (Pierce) overnight at 4 degrees and washed 5 times with RIPA buffer and then eluted with protein loading buffer. Eluates were analyzed by Western blot.

NFAT activation assay.
NFAT (Nuclear factor of activated T-cell) activation assay was performed with a GFP reporter as described previously . The 2B4 reporter cell line was transfected with TREM2-IRES-DAP12 or TREM R47H-IRES-DAP12 were incubated with Aβ42 oligomers for 12 hours. NFAT signaling activation was measured by the expression of GFP. The expression of GFP was analyzed by flow cytometry and reporter activity is presented as percentage of cells which express GFP.

Statistical analysis
Statistics were performed with GraphPad Prims ( Triggering receptor expressed on myeloid cells 2 (TREM2) is a member of the immunoglobulin superfamily highly expressed in microglial cells. Rare coding variants of TREM2 (R47H, R62H) are associated with increased risk for Alzheimer's disease (AD), but how they confer this risk remains uncertain.

Results
We assessed binding of TREM2, AD associated TREM2 variants and select TREM family members to various forms of Aβ and APOE in multiple assays. TREM2 interacts directly with various forms of Aβ in multiple assays, with highest affinity interactions observed between TREM2 and soluble Aβ42 oligomers.
AD-associated TREM2 variants bind various forms of Aβ with similar affinity as wild type TREM2. High affinity binding of TREM2 to Aβ oligomers (KD < 1e-12) is characterized by very slow dissociation.
Preincubation with Aβ oligomers is shown to block the lower affinity interaction of APOE. Our data show that while mouse Trem2, Trem1 and Treml1 show the same high affinity interaction with Aβ oligomers as human TREM2. In cellular assays, AD associated variants of TREM2 reduced the amount of Aβ42 internalized, and in NFAT reporter assay the R47H variant decreased NFAT signaling activity in response to Aβ42.

Impact
These studies identify a high affinity interaction between TREM2 and Aβ that can block interaction with another ligand, APOE. These data link TREM2 to physical interactions with two key molecules in Alzheimer's disease, A and APOE, and suggest that partial loss of function of TREM2 mediates increased risk for AD.

Figure 3. Binding of Aβ42 oligomers on soluble TREM-Fc prevent binding of APOE3 in BioLayer Interferometry.
BioLayer Interferometry experiments were performed to evaluate the binding capacity of APOE3 on soluble TREM2-Fc pre incubated with Aβ42 oligomers. Schematic representations of two sequential associations of ligand on soluble TREM2-Fc are represented by I and II. Soluble TREM2-Fc loaded on a sensor is first incubated with APOE3 (association I) followed by a second incubation with Aβ42 oligomers (association II) or the reverse-incubated with Aβ42 oligomers (association I) followed by a second incubation with APOE3 (association II). All experiments were replicated 3 independent times. ELISA plates were coated with purified soluble TREM2-Fc, and then incubated with various concentrations of Aβ42 monomers or 1 µM Aβ42 oligomers.
B-HEK293T cells transfected with TREM2 or AD variant incubated with 500 nM Aβ42 monomers were lysed for Aβ42 quantifications by ELISA. Expression of TREM2 increases Aβ42 level detected compared to mock control as well the AD variant. Results were averaged ± standard error (n = 9, # p < 0.05 compared with TREM2, ## p< 0.01 compared with TREM2, *** p< 0.001 compared with mock, Repeated Measures ANOVA, Tukey's Multiple Comparison Test).
C-HEK293T cells transfected with TREM2 were incubated with Aβ42 oligomers for 16 hours. Conditioned media and cell lysate were analyzed by Western blot for TREM2 (RIPA lysate) and soluble TREM2 (media) detection. Experiment was replicated 3 independent times.
D-HEK293T cells transfected with TREM2 were subjected to cell membrane biotinylation for purification. Bottom panel in D shows the presence of TREM2 in the purified product. Endogenous APP and β-tubulin were used as controls for the purification (middle and upper panels). Experiment was replicated 3 independent times A-Schematic representation an Aβ antibody association on soluble TREM2-Fc pre incubated with Aβ42 oligomers.
B-Western blot analysis of purified soluble TREM-Fc and soluble trem2-Fc family members and Aβ oligomers.