A split-luciferase complementation, real-time reporting assay enables monitoring of the disease-associated transmembrane protein TREM2 in live cells

Triggering receptor expressed on myeloid cells 2 (TREM2) is a single transmembrane molecule uniquely expressed in microglia. TREM2 mutations are genetically linked to Nasu-Hakola disease and associated with multiple neurodegenerative disorders, including Alzheimer's disease. TREM2 may regulate microglial inflammation and phagocytosis through coupling to the adaptor protein TYRO protein-tyrosine kinase-binding protein (TYROBP). However, there is no functional system for monitoring this protein-protein interaction. We developed a luciferase-based modality for real-time monitoring of TREM2-TYROBP coupling in live cells that utilizes split-luciferase complementation technology based on TREM2 and TYROBP fusion to the C- or N-terminal portion of the Renilla luciferase gene. Transient transfection of human embryonic kidney 293 cells with this reporter vector increased luciferase activity upon stimulation with an anti-TREM2 antibody, which induces their homodimerization. This was confirmed by ELISA-based analysis of the TREM2-TYROBP interaction. Antibody-mediated TREM2 stimulation enhanced spleen tyrosine kinase (SYK) activity and uptake of Staphylococcus aureus in microglial cell line BV-2 in a kinase-dependent manner. Interestingly, the TREM2 T66M mutation significantly enhanced luciferase activity without stimulation, indicating constitutive coupling to TYROBP. Finally, flow cytometry analyses indicated significantly lower surface expression of T66M TREM2 variant than wild type or other TREM2 variants. These results demonstrate that our TREM2 reporter vector is a novel tool for monitoring the TREM2-TYROBP interaction in real time.

Triggering receptor expressed on myeloid cells 2 (TREM2) 2 is expressed by cells of the myeloid lineage (1), especially microglia (2,3). TREM2 is mostly found in intracellular stores but can rapidly translocate to the cell surface upon exocytotic stimulation (4). The adaptor protein TYRO protein-tyrosine kinasebinding protein (TYROBP; also known as DAP12) is an intracellular signaling molecule for TREM2 signaling (for reviews, see Refs. 2 and 5). Upon receptor ligation, tyrosine residues within the TYROBP immunoreceptor tyrosine-based activation motif region undergo phosphorylation by Src kinases. This leads to inflammatory signaling and reorganization of the actin cytoskeleton via activation of small guanine nucleotide-binding proteins such as Rac and Cdc42 for endocytic functions (6).
The clinical significance of TREM2 and TYROBP was first highlighted by their mutations in a rare autosomal recessive inflammatory neurodegenerative disorder, Nasu-Hakola disease. Early studies revealed that Nasu-Hakola disease patients carried loss-of-function mutations in TYROBP or TREM2 (7)(8)(9). The presence of these mutations results in exaggerated responses to Toll-like receptor activation and elevated levels of proinflammatory cytokines in plasma and brain (10 -12). More recent studies revealed that specific variants in TREM2 were associated with late-onset AD, Parkinson's disease, amyotrophic lateral sclerosis (13), and frontotemporal dementia (FTD) (13)(14)(15)(16)(17)(18)(19). The TREM2 R47H variant was associated with AD and FTD (15,16), whereas the T66M variant was associated with FTD or FTD-like syndrome (19), and the S116C variant was found in one case of FTD (19).
It is unclear how these mutations affect the development of neurodegenerative diseases. Recent in vitro data suggest that intracellular TREM2 is processed into a soluble form (sTREM2), and this form may have biological functions (20). Disease-associated TREM2 mutations may result in deficient processing and secretion of sTREM2 that lead to an accumula-tion of immature TREM2 in the cytoplasm (20). Although its biological function is unknown, sTREM2 is detected in the cerebrospinal fluid (CSF) and plasma of healthy normal patients. The level of sTREM2 in CSF was significantly lower in AD and FTD patients of one study (20), whereas other studies reported it to be elevated (21,22).
The role of TREM2 in amyloid clearance has been under intense investigation. APPϩPS1 mice expressing familial ADlinked variants of amyloid precursor protein (13) and presenilin-1 (PS1) show reduced accumulation of amyloid-␤ peptide (A␤) by disruption of TREM2 (23), whereas others report enhanced A␤ clearance in 5XFAD mice lacking TREM2 (24). These apparently contrasting results suggest that TREM2 function can be affected by mutated PS1 and should be investigated in more physiological models.
The disease-linked TREM2 mutations affect phagocytic function of myeloid cells. Phagocytosis of A␤ was impaired due to expression of R47H or T66M TREM2 mutation (20). Patients with disease-associated TREM2 variants may then have a build-up of A␤ and apoptotic debris due to deficient processing of TREM2 and impaired phagocytosis. To investigate TREM2 and its disease-associated mutations, we developed a bioluminescence-based assay that allows us to monitor real-time TREM2 coupling to TYROBP in vitro with a construct that utilizes split-luciferase complementation technology.

TREM2-CLuc-IRES-TYROBP-NLuc construct complements luciferase activity in response to anti-TREM2 antibody
For examination of the real-time protein-protein interaction between TREM2 and TYROBP, we developed a mammalian expression vector (TREM2-CLuc-IRES-TYROBP-NLuc) that utilizes the "bait-prey" split-Renilla luciferase complementation assay (Fig. 1A) (25)(26)(27). Renilla luciferase catalyzes the oxidation of coelenterazine, a luciferin, to emit light (28). The amount of light emitted is measured as relative light units (RLU) and can be directly correlated to the magnitude of protein-protein interaction (Fig. 1B). The TREM2-CLuc-IRES-TYROBP-NLuc plasmid was transfected into human embryonic kidney (HEK) 293 cells followed by stimulation with specific ligands. Luciferase activity was measured over 30 min, and area under the curve (AUC) was calculated. HEK293 cells transfected with TREM2-CLuc-IRES-TYROBP-NLuc showed some basal level of luciferase activity that may represent spontaneous TREM2 coupling to TYROBP that occurs independently of ligand, although this was not significantly higher than background (Fig. 2, A and B). As expected, our TREM2-CLuc-TYROBP negative control construct elicited luciferase activity that was also no higher than background (Fig. 2, A and B). It has previously been shown that a TREM2-cross-linked antibody can activate TREM2 (29). Indeed, monoclonal anti-TREM2 antibody stimulation at 10 and 20 g/ml induced significantly higher luciferase activity up to a 3-fold net increase over time compared with unstimulated transfected cells (Fig. 2, E and F). Treatment of transfected cells with 20 g/ml antibody did not induce a significantly higher effect compared with 10 g/ml, suggesting its saturated binding to TREM2. Treatment of trans-fected cells with control IgG resulted in only a maximum of about a 0.3-fold net increase at 1 g/ml (supplemental Fig. S1, A and B). Although this difference was statistically significant, there is a greater biological change upon stimulation with anti-TREM2 antibody. Research suggests that TREM2 activation may be increased in response to the presence of A␤ because its expression is increased in amyloid plaque-associated microglia in the APP mouse brain (30). However, there was no significant effect of A␤42 stimulation (Fig. 2, C and D) or anti-TYROBP antibody (Fig. 2, G and H) even at the highest concentrations. These data demonstrate that our TREM2-CLuc-IRES-TY-ROBP-NLuc construct is capable of being utilized to measure the TREM2-TYROBP interaction by ligand stimulation.

Anti-TREM2 antibody induces TREM2 coupling to TYROBP in HEK293 cells
To support our luciferase data, we transfected HEK293 cells with TREM2-CLuc-IRES-TYROBP-NLuc and stimulated with anti-TREM2 antibody followed by quantification of the TREM2-TYROBP interaction by sandwich ELISA (Fig. 3A). For this assay, treatment with the tyrosine phosphatase inhibitor sodium orthovanadate and cross-linker dimethyl 3,3Ј-dithiobispropionimidate (DTBP) was necessary. This is because TREM2 signaling presumably is dependent on tyrosine phosphorylation of TYROBP as well as the transient nature of their interaction due to the secondary association of TYROBP with SHIP1 (31,32). Cells stimulated with anti-TREM2 antibody showed a significant increase in TREM2-TYROBP interactions versus unstimulated cells, suggesting that increased stabilization of TREM2 coupling to TYROBP was due to antibody treatment (Fig. 3B).

Anti-TREM2 antibody stimulation enhances phagocytosis of Staphylococcus aureus in BV-2 microglial cells
To further investigate whether the TREM2-TYROBP interaction facilitates phagocytic activity, we performed a phagocytosis assay with the BV-2 mouse microglial cell line and S. aureus particles conjugated with Alexa Fluor 594 and fluorescence-activated cell sorting (FACS). BV-2 cells endogenously express TREM2 and TYROBP (33,34). We confirmed the endogenous distribution of TREM2 by flow cytometry and found that 25.3% of TREM2 is expressed on the cell surface (supplemental Fig. S2). We stimulated BV-2 cells with anti-TREM2 antibody to induce endogenous TREM2-TYROBP coupling. Cells stimulated with anti-TREM2 antibody were labeled with a donkey anti-goat Alexa Fluor 488-conjugated secondary antibody. Gating of cells was done using unstained cells or cells positive for Alexa Fluor 488 or 594 alone (Fig. 4, A-D). The total percentage of phagocytic cells is indicated in each graph (Fig. 4, E-I). The S. aureus Alexa Fluor 594 ϩ population was then separated into phagocytic high (Hi) versus low (Lo) groups (Fig. 4, E-G), and phagocytosis was determined by

Reporter analysis of TREM2-TYROBP signaling
analyzing the percentage of cells in each group over the total cell population (Fig. 4H). Untreated cells had a significantly lower percentage of phagocytic Hi (3.9%) populations versus phagocytic Lo (11.2%) populations. Treatment of microglia with anti-TREM2 antibody significantly increased both the percentage of phagocytic Hi (by 60%) and Lo populations (by 138.1%) as compared with untreated cells. In control experiments, treatment of BV-2 cells with control IgG did not significantly increase low and high phagocytosis-dependent mechanisms. A greater proportion of cells were in the high-uptake population after treatment with anti-TREM2 antibody as compared with the IgG group, indicating that activation of TREM2 is more responsible for increasing the phagocytic Hi population. These findings are substantiated by the use of TREM2specific inhibitor to spleen tyrosine kinase (SYK) (Fig. 4H). Control IgG-stimulated phagocytosis was insensitive to SKY inhibition in low-phagocytosis cell populations. TREM2-dependent mechanisms were sensitive to treatment with the SYK inhibitor in both phagocytic Lo (Fig. 4H) and phagocytic Hi populations (Fig. 4H). To explore whether the observed trend toward increased phagocytosis in control IgG-treated BV2 cells is accomplished through Fc receptors, the normal goat F(abЈ) 2 fragment was also tested; it failed to increase the population of low-or high-phagocytic groups (Fig. 4H). These data indicate that enhancing the TREM2-TYROBP interaction by anti-TREM2 antibody stimulation increases the phagocytic Hi population, which biologically links the split-luciferase activity to phagocytic functions of myeloid cells.

TREM2 induces activation of SYK
Coupling of TREM2 with TYROBP activates intracellular signaling via recruitment of SYK. To directly demonstrate the intracellular signaling of TREM2 after antibody stimulation, we stimulated the parental BV-2 microglial cell line with anti-TREM2 or nonspecific IgG in the presence or absence of a SYK inhibitor (CAS 622387-85-3) and immunoprecipitated SYK from the cell lysate. The immunoprecipitate was then used in an in vitro SYK activity assay that measured SYK's functional conversion of ATP to ADP (Fig. 5). Treatment with SYK inhibitor alone reveals background kinase activity in the immunoprecipitants, but stimulation of BV-2 cells with anti-TREM2 antibody induced significantly higher activation of SYK in the absence of SYK inhibitor. Control IgG stimulation had no effect on SYK activity. These data further illustrate that anti-TREM2 antibody stimulation can activate specific TREM2 intracellular signaling pathways.

T66M mutation of TREM2 induces enhanced basal luciferase activity
We next examined the effect of disease-associated TREM2 mutation of TYROBP coupling in our system. For that purpose, TREM2 variants associated with AD (R47H), Nasu-Hakola disease and FTD (T66M), and a case of FTD (S116C) were inserted into the TREM2-CLuc-IRES-TYROBP-NLuc vector. Immunostaining of permeabilized cells showed mostly similar distribution of TREM2 and TYROBP across all groups, indicative of their intracellular interactions (supplemental Fig. S3A). Transfection of HEK293 cells with wild-type (WT) TREM2 and its variants followed by ELISA for protein levels of TREM2 (supplemental Fig. S3B) and TYROBP (supplemental Fig. S3C) shows no differences in expression efficiency. On average, the ratio of TREM2/TYROBP expression was around 1.0 for each variant, indicating no difference in expression (supplemental Fig. S3D). The effect of anti-TREM2 antibody stimulation on . Quantification of TREM2⅐TYROBP complex by "sandwich" ELISA. HEK293 cells were transfected with TREM2-CLuc-IRES-TYROBP-NLuc followed by anti-TREM2 antibody (Ab) stimulation to induce TREM2 coupling to TYROBP and protein cross-linking with DTBP and phosphatase inhibitor sodium pervanadate (Na 3 O 4 V) (A). Following cell lysis, protein was incubated in anti-TYROBP-coated wells and labeled for TREM2 with biotinylated anti-TREM2⅐HRPconjugated streptavidin complex for quantification of TREM2⅐TYROBP protein complexes. Average (Ave) OD was quantified by ELISA. Data represent the arithmetic difference between a given treatment condition and the average OD of untransfected controls (n ϭ 4 per group) (B). Significant differences were determined by one-way ANOVA followed by Tukey's post-test. Error bars represent S.D. † † and † † † denote p Ͻ 0.01 and p Ͻ 0.001, respectively, versus untransfected cells. ### denotes p Ͻ 0.001 versus unstimulated cells transfected with TREM2-TYROBP. ** and *** denote p Ͻ 0.01 and 0.001, respectively, versus those indicated.
the coupling of TREM2 variants with TYROBP was monitored by real-time luciferase activity (Fig. 6, A, C, and E) and AUC ( Fig. 6, B, D, and F). Interestingly, the T66M mutation elicited higher luciferase activity without stimulation and did not show enhanced luciferase activity by anti-TREM2 antibody stimulation, showing the autonomous coupling of the T66M variant with TYROBP in response to anti-TREM2 antibody (Fig. 6, C and D). Unlike WT TREM2, variant TREM2 constructs did not show significant differences in luciferase activity in response to increasing antibody doses. Control experiments demonstrate that the stimulating effect of anti-TREM2 antibody is specific because treatment with control IgG induced no change in any of the TREM2 variants (supplemental Fig. S1, C-H).

T66M TREM2 variant shows reduced cell surface expression of TREM2
We also determined whether there was a difference in cell surface expression compared with WT TREM2. For this purpose, the cell surface TREM2 ϩ populations were assessed by FACS analysis of TREM2-phycoerythrin (PE)-stained HEK293 cells that were untransfected or transfected with WT TREM2 or R47H, S116C, or T66M variant (Fig. 7, A and B). TREM2 is mostly localized in the cytoplasm, which is reflected by the low number of cell surface TREM2 ϩ cells (8.67% in TREM2 WT, 6.37% in R47H, 6.11% in S116C, and 2.77% in T66M group). The number of cell surface TREM2 ϩ cells was less than 50% of other groups. To establish that the differences induced by variant TREM2 constructs were not due to differences in transfec-

Reporter analysis of TREM2-TYROBP signaling
tion efficiency, the cells were also stained with TREM2-PE after permeabilization, which showed similar numbers of TREM2 ϩ cells (30.3% in TREM2 WT, 28.1% in R47H, 25.6% in S116C, and 25.8% in T66M; Fig. 7, C and D). Mean fluorescence intensity (MFI) of gated TREM2 cells from the roughly 10,000 cells was significantly higher in the TREM2 WT and S116C groups versus untransfected groups, whereas R47H and T66M groups showed significantly lower MFIs of gated TREM2 ϩ cells versus the TREM2 WT group (Fig. 7E). There was no difference in the MFI of TREM2 ϩ cells in permeabilized cells (Fig. 7F), demonstrating similar expression of total TREM2 in transfected cells. Taken together, these data suggest that the T66M mutation induces constitutive coupling with TYROBP as evident with the higher split-luciferase activity, which may prevent its cell surface expression.

Discussion
TREM2 has both neuropathological and genetic associations with AD. However, how it affects onset or progression of disease is unclear due to inconsistent study results. In primary microglial cells, TREM2 promoted phagocytosis of apoptotic neurons without stimulating cytokine production (35). This suggests that disease-associated mutations in TREM2 may impair phagocytosis of apoptotic neurons and increase the release of proinflammatory cytokines. Microglia surrounding amyloid plaques show elevated expression of TREM2 in AD (3,36,37). A recent examination of AD-like transgenic mice and post-mortem human AD brains revealed that these TREM2expressing microglia tightly surround amyloid fibrils to promote their compaction, and TREM2 haplodeficiency or expression of the R47H polymorphism correlates with a reduction in this plaque compaction and increased axonal dystrophy in the regions surrounding plaques (37). Levels of TREM2 were also significantly elevated in the hippocampi of aged APPϩPS1 mice, and overexpression of TREM2 in primary microglia from these mice facilitated phagocytosis of A␤42 in vitro (38). In support of these findings, a recent study found that APPϩPS1 mice deficient in TREM2 had reduced association of macro-phages with A␤ plaques and reduced expression of proinflammatory cytokines such as interleukin-6 (IL-6) and IL-1␤ (23). These mice also had reduced levels of amyloid deposition in the hippocampi compared with APPϩPS1/TREM2 ϩ/ϩ controls (23) despite the predicted reduction in phagocytic capability of macrophages with deficient TREM2 expression. These studies suggest that TREM2 is induced by the interaction of microglia with amyloid plaques and potentially facilitates their clearance while reducing the amount of damage done to neighboring neurons. These findings are contradictory to an earlier study showing that 5XFAD mice lacking TREM2 have enhanced accumulation of microglia around amyloid plaques (24). These contradictory findings suggest that TREM2 may function differently in mouse versus human brains and during different disease stages and that both enhancers and inhibitors of TREM2 signaling should be investigated to understand the species-specific difference in TREM2 function.
In this study, we successfully developed a split luciferasebased monitoring system in live cells to allow for real-time analysis of TREM2 coupling to TYROBP using a bioluminescence-based assay. Our TREM2-CLuc-IRES-TYROBP-NLuc construct shows a dose-dependent increase in luciferase activity in response to monoclonal anti-TREM2 antibody stimulation, indicative of TREM2-TYROBP binding. We used this construct to study the effects of AD-and FTD-associated TREM2 mutations on the TREM2-TYROBP interaction. We show that a TREM2-specific monoclonal antibody can stimulate TREM2-TYROBP coupling in a dose-dependent manner. Our data, however, do not show positive stimulation of TREM2-TYROBP coupling by human A␤42 peptides, which are mainly dimers. This suggests that A␤ stimulation of microglia induces TREM2 expression in the AD brain even though it may not be the ligand of TREM2. A recent study demonstrated that microglia are more efficient at taking up A␤ when it forms a complex with a lipoprotein, such as low-density lipoprotein, compared with A␤ by itself (39). The degree of A␤ uptake is also dependent on the degree of TREM2 expression in microglia (39). However, more detailed examinations are necessary to determine whether the TREM2-A␤ interaction is specific to conformations, such as oligomers, protofibrils, or fibrils.
Microglia show increased phagocytosis of apoptotic neurons when coated with apoE (40), suggesting that TREM2 may function to activate microglial phagocytosis of apoE-coated cells that are undergoing apoptosis. Phosphatidylserine is first detected by milk fat globule-EGF factor 8 (41,42), which is secreted by microglia and can bind to integrin receptors (␣ v ␤ 3 or ␣ v ␤ 5 ) on the cell surface (43). In addition, phosphatidylserine on the apoptotic body is also detected by growth arrest-specific gene 6 and protein S, which stimulate the Tyro3, Axl, and Mertk subfamily of receptor tyrosine kinases for the immune response and phagocytosis (44 -46). The interaction of TREM2 with apoptotic bodies may also require a bridging molecule similar to milk fat globule-EGF factor 8, growth arrest-specific gene 6, or protein S, which are not secreted by HEK293 cells. A more physiological model will be necessary to complement the missing molecule for the development of an apoptotic body-stimulated TREM2 reporter system. We also demonstrated that stimulation of microglial cells with an anti-TREM2 antibody enhances their ability to phagocytose S. aureus, but not yeast-derived zymosan, by BV-2 cells. This is consistent with the previous report showing the TREM2-dependent uptake of bacteria (S. aureus, Escherichia coli, and Francisella tularensis) but not zymosan (47). Although it is still unclear which bacterial molecules are the binding targets of TREM2, carbohydrates and lipopeptides are likely good candidates. This is a convenient alternative measure to assess the effect of TREM2 activation on phagocytosis in the murine microglial cell line. This assay can be utilized in conjunction with TREM2 variant cell lines to further examine how polymorphisms in TREM2 can affect TREM2-TYROBP signaling and phagocytic capability in vitro. The sensitivity of low-and highphagocytic TREM2-stimulated cell populations to SYK inhibition also demonstrates the activation of SYK-dependent phagocytic mechanisms. Although treatment with control IgG was able to stimulate phagocytosis, it was insensitive to SYK inhibition, demonstrating that not all Fc receptors are SYK-dependent. Potent activation of SYK by TREM2 supports these data and illustrates a TREM2-SYK-dependent mechanism for phagocytosis by microglial cells.
Our split-luciferase system is of great value to further examine the effects of TREM2 polymorphisms in AD as these have the greatest association with risk of AD after the well known genetic risk factor apoE. We demonstrate a significant effect of the T66M variant on constitutive TREM2 coupling to TYROBP. This was unexpected because the homozygous T66M TREM2 mutation was previously identified in Nasu-Hakola disease and predicted to be a loss-of-function mutation (9). Other Nasu-Hakola disease-linked mutations of TREM2 or TYROBP consistently cause loss-of-function mutations (48). Our data show that the T66M mutation is a gain-of-function mutation in terms of its coupling to TYROBP. However, our FACS-based analysis of the cell surface expression TREM2 variants shows that the T66M variant is poorly expressed on the cell surface compared with other TREM2 variants. TREM2 can be expressed on the cell surface by ionomycin treatment-induced Ca 2ϩ influx and exocytotic stimulation (4). We tested ionomycin treatment of HEK293 cells after transient transfection of TREM2, but it was unchanged (data not shown). This T66M mutation of TREM2 may cause constitutive coupling with TREM2, which induces SYK activation and endocytosis of the complex, suggesting that T66M TREM2 is mostly retained

Reporter analysis of TREM2-TYROBP signaling
intracellularly and does not engage in phagocytic function on the surface. Thus, we speculate that T66M is a gain-of-coupling mutation but consequently a loss-of-phagocytic-function mutation. Our finding on the T66M TREM2 variant also replicates those found in Kleinberger et al. (20) and Song et al. (49) who show reduced cell surface expression of TREM2 carrying the T66M mutation. Dysfunctional processing of T66M TREM2 resulted in an accumulation of immature TREM2 and reduced production of plasma membrane-bound or sTREM2 in cell medium in vitro, which supported their discovery that AD patients show reduced levels of sTREM2 in the CSF (20). However, these findings are contrary to a more recent report that AD patients had significantly higher sTREM2 concentrations in the CSF that correlated with markers of neurodegeneration and glial activation (22). Further study is needed to support these findings. The T66M mutation may result in hyperactive TREM2-TYROBP signaling and impaired processing or a TREM2 that is detrimental to immune homeostasis and phagocytosis, thus increasing the risk for AD.
In summary, we demonstrated that expression of the T66M mutation in HEK293 cells resulted in constitutive TREM2 activation and coupling to TYROBP. Thus, the T66M TREM2-CLuc-IRES-TYROBP-NLuc construct would be a good platform to screen inhibitors of TREM2-TYROBP coupling. Our TREM2-CLuc-IRES-TYROBP-NLuc construct would be more suitable for screening enhancers for TREM2-TYROBP coupling. These reporter constructs and complementing phagocytosis and cell surface expression assays may be useful for identifying ligands that can restore normal TREM2 functioning and coupling to TYROBP in patients with disease-associated polymorphisms in TREM2.

Plasmid design for TREM2-CLuc-IRES-TYROBP-NLuc constructs
We chose an internal ribosome entry site (IRES) backbone vector (pIRES; Clontech) under the control of the CMV immediate early promoter to allow for two cloning sites to express both human TREM2 and TYROBP genes as split-luciferase fusion proteins from a single mRNA (Fig. 1A). All restriction sites are unique for each insert. All amplifications by PCR were done using Pfu UltraII Hot start PCR Master Mix (Agilent Technologies, Santa Clara, CA). Human TREM2 (GenBank TM accession number NM_018965.2; 693 bp) and TYROBP cDNA sequences (GenBank accession number BT009851.1; 342 bp) were amplified by PCR from the human brain-derived cDNA library as a template using primer sets as listed in Table 1. The C terminus of the Renilla luciferase gene (CLuc) was amplified by PCR from pDuex-Bait vector, and the N terminus of the Renilla luciferase gene (NLuc) was amplified from pDuex-Prey vector using the primer sequences given in Table 1 (26). All primers were custom synthesized by Invitrogen. The PCR amplicons were purified after gel electrophoresis using the QIAquick Gel Extraction kit (Qiagen, Valencia, CA) and ligated into the TOPO TA cloning pCR2.1 vector (Invitrogen). After confirmation of the DNA sequences of TREM2, TYROBP, CLuc, and NLuc DNA fragments, the plasmids were digested with restriction enzymes from New England Biolabs (M0202; Ipswich, MA) as indicated in Fig. 1A, purified after gel electrophoresis using QIAquick Gel Extraction kit, and sequentially inserted into the pIRES vector to develop the TREM2-CLuc-IRES-TYROBP-NLuc plasmid. Point mutations for R47H (16,19), T66M (19), and S116C of human TREM2 (19) were designed and incorporated into the vector by GenScript (Piscataway, NJ). DNA plasmids were purified using the Endo-Free Plasmid Maxi kit for transfection studies (Qiagen).

Reporter analysis of TREM2-TYROBP signaling
injection at 0.5-s integration before injecting substrate into the next well. The luminescence was measured from each individual well in succession at 0.5-s integration repeatedly over 30 min.

TREM2-TYROBP cross-linking assay
To confirm that luciferase activity measured in our luminescence assays was due to a physical interaction between TREM2 and TYROBP, we performed a cross-linking assay with protein quantification by sandwich ELISA. Fig. 3A illustrates the experimental scheme used for measuring concentration of TREM2⅐TYROBP protein complexes. HEK293 cells were plated in a 100-mm dish at 70% confluence before transfection with 10 g of TREM2-CLuc-IRES-TYROBP-NLuc with Lipofectamine 2000. Cells were then stimulated with 20 g/ml anti-TREM2 antibody (goat polyclonal IgG, AF1828, R&D Systems) for 20 min in DMEM to induce TREM2 coupling to TYROBP. To stabilize the TREM2-TYROBP interaction for measurement by ELISA, we utilized adapted protocols to induce protein cross-linking (31,50,51). Briefly, we replaced media with DMEM containing 4 mM DTBP (D2388, Sigma-Aldrich), an amidine protein cross-linker with an 11.9-Å spacer arm, and 10 mM phosphatase inhibitor sodium vanadate with 0.6% hydrogen peroxide (sodium pervanadate; Sigma-Aldrich) for 10 min at 37°C. Sodium vanadate was necessary to detect the interaction of TREM2 and tyrosine-phosphorylated TYROBP, which would be inactivated by the association of tyrosine phosphatase SHIP1 (31,32). After incubation, media were aspirated, and cells were washed with PBS before lysing in radioimmune precipitation assay buffer (50 mM Tris hydrochloride, 150 mM sodium chloride, 1 mM EDTA, 0.1% sodium dodecyl sulfate, 0.5% sodium deoxycholate, and 1% Nonidet P-40) containing 1% protease inhibitor phenylmethylsulfonyl fluoride and 1 mM sodium pervanadate for 1 h on ice. Tris quenches cross-linking by any remaining DTBP. The lysate was centrifuged at 14,000 rpm for 30 min at 4°C. Protein was quantified using the BCA Protein Assay kit (Pierce) according to the manufacturer's instructions.

TREM2-TYROBP sandwich ELISA
Tissue culture-grade 96-well plates were coated with 50 l of anti-TYROBP antibody (rabbit polyclonal IgG, sc-20783) at a concentration of 1 g/ml in coating buffer (0.2 M sodium bicarbonate, pH 9.4) overnight at 4°C. Coated wells were washed and then blocked with 10% normal goat serum in PBS for 1 h at room temperature. HEK293 protein lysates from the TREM2-TYROBP cross-linking assay were applied to TYROBP-coated wells at 100 g in 100 l of radioimmune precipitation assay buffer and incubated overnight at 4°C. To label TYROBP protein cross-linked with TREM2, TYROBP-bound protein was labeled with a biotinylated anti-human TREM2 antibody (goat polyclonal IgG, BAF1828, R&D Systems)⅐HRP-conjugated streptavidin (21126, Life Technologies) complex (0.5 g/ml each) in 0.1% Triton X-100 in PBS for 1 h at room temperature. Wells were then incubated with 100 l of a stabilized chromogen, tetramethylbenzidine, for 30 min in the dark at room temperature followed by addition of 100 l of Stop solution (Life Technologies). Optical density (OD) was read at 450 nm with a SpectraMax M2 microplate reader (Molecular Devices, Sunnyvale, CA), and results were quantified by the arithmetic difference between the given treatment and the average OD of untransfected controls for n ϭ 4 per group.

Phagocytosis assay
To determine whether activation of TREM2 enhanced the ability of BV-2 cells to phagocytose, BV-2 microglial cells were plated at 1.0 ϫ 10 6 /well in a 6-well plate. Cells were then serumstarved overnight before the phagocytosis assay. Some cells were pretreated with 14 M SYK inhibitor (CAS 622387-85-3, EMD Millipore, Billerica, MA) for 30 min to test TREM2specific inhibition of phagocytosis. Alexa Fluor 594-tagged S. aureus particles (S-23372, Life Technologies) were prepared beforehand by centrifugation at 10,000 ϫ g followed by resuspension in DMEM and 15-min sonication in a water bath sonicator. BV-2 cells were then stimulated with 20 g/ml antihuman TREM2 antibody (goat polyclonal IgG, BAF1828), goat IgG control, or goat F(abЈ) 2 fragment control (ChromPure 005-000-006, Jackson ImmunoResearch Laboratories, West Grove, PA) and a 1:5 ratio of cells:S. aureus particles for 30 min. Cells were trypsinized immediately after completion of the assay to generate a single cell suspension and remove non-phagocytosed S. aureus and prepared for analysis of phagocytic cells in technical duplicate by flow cytometry. Blocking of nonspecific Fc-mediated interactions was done with 4 l of anti-mouse CD16/CD32/100 l for 20 min on ice. To label cells treated with anti-TREM2 antibody, cells were incubated with 5 g/ml donkey anti-goat Alexa Fluor 488 secondary antibody (A11055, Invitrogen) for 30 min. Cells not treated with Alexa Fluor 594tagged S. aureus particles or Alexa Fluor 488 secondary antibody were used to gate cells that were positive for Alexa Fluor 488 or 594 signal. Cells were examined by flow cytometry within 1 h of completing the staining procedure using a BD LSR II instrument with BD FACSDiva 6.2.1 software (BD Biosciences) and analyzed with FlowJo software (FlowJo LLC, Ashland, OR).

SYK activity assay
Parental BV-2 cells were stimulated in suspension with either anti-TREM2 antibody (BAF1828) or nonspecific IgG control (AB-108-C, R&D Systems) at a concentration of 5 g/ml each in Opti-MEM at 37°C for 30 min. SYK inhibitor (CAS 622387-85-3) was also administered in tandem at a concentration of 14 M. Cells were lysed on ice in lysis buffer (25 mM Tris-HCl, pH 7.4, Triton X-100, 150 mM NaCl, 1 g/ml aprotinin, 10 g/ml leupeptin, 1 mM EDTA, 50 mM NaF, and 1 mM sodium orthovanadate), and cell lysate was extracted and incubated with anti-SYK rabbit polyclonal antibody (sc-573, Santa Cruz Biotechnology) for 3 h. Immunoprecipitation was then conducted with 20 l of protein A-agarose beads (ab193254, Abcam). Next SYK was eluted from the beads with 0.1 M glycine for 10 min and stabilized with 50 mM Tris. SYK extracts were then assessed for levels of activity using an ADP-Glo assay kit (V6930, Promega) according to the manufacturer's procedure.
Cells were analyzed for cell surface or intracellular expression of TREM2 and TYROBP. For analysis of intracellular TREM2 and TYROBP, cells were fixed in 4% paraformaldehyde followed by permeabilization with 0.5% Tween 20 in PBS containing 1% bovine serum albumin. Cells used for cell surface expression of protein remained unfixed and unpermeabilized to maintain the integrity of the cell membrane. Cells were then incubated with anti-TYROBP antibody (rabbit polyclonal IgG, sc-20783) and PE-conjugated anti-human/mouse TREM2 (rat monoclonal IgG2b clone 237920, FAB17291P) for 1 h. This was followed by 30-min secondary antibody incubation with donkey anti-rabbit conjugated to Alexa Fluor 647 (A-31573, Invitrogen) at 10 g/ml for fluorescence labeling of TYROBP. At least 10,000 cells were analyzed by flow cytometry using a BD LSR II instrument with BD FACSDiva 6.2.1 software and FlowJo software. Gating was aided by analysis of single positive PE-TREM2-only and TYROBP-Alexa Fluor 647-only cells as well as negative control cells that were treated with neither anti-TREM2 nor anti-TYROBP antibody.

Immunofluorescence
HEK293 cells (1 ϫ 10 5 cells/well) were plated on poly-Dlysine-coated coverslips in 24-well plates, transfected with 800 ng of TREM2-CLuc-IRES-TYROBP-NLuc (WT TREM2) or R47H, T66M, or S116C variant constructs as described under "Flow cytometry for TREM2/TYROBP expression" and 48 h post-transfection fixed by 4% paraformaldehyde in PBS for 20 min. For the cell surface detection of TREM2, cells were blocked by 5% donkey serum in PBS for 30 min and stained with 1:200 anti-TREM2 rat monoclonal antibody as described above followed by washing and permeabilization of the cells with 0.5% Triton X-100 and blocking with 5% donkey serum prior to staining with 1:1000 anti-TYROBP rabbit polyclonal antibody as described above. For total TREM2 staining, cells were first permeabilized with 0.5% Triton X-100 and blocked with 5% goat serum in PBS for 30 min followed by staining with anti-TREM2 rat and anti-TYROBP rabbit antibodies in a similar manner. After washing, the bound primary antibodies were detected by 1:500 Alexa Fluor 594-conjugated donkey anti-rat and Alexa Fluor 488-conjugated donkey anti-rabbit antibodies (Invitrogen), respectively. The immunostained cells on the coverslips were mounted on Superfrost Plus slides (Fisher Scientific) with Fluoromount mounting medium (Sigma-Aldrich), and images were captured using a Nikon Eclipse TE-2000U with attached charge-coupled device camera with Nikon CFI Plan Fluor 60ϫ oil objective.

Statistics
All data are presented as mean values Ϯ S.D. For comparison between two groups, data were analyzed by Student's t test. For multiple mean comparisons, data were analyzed by one or twoway ANOVA followed by Tukey or Bonferroni post hoc analysis using statistics software (Prism 4.0, GraphPad Software, Inc., San Diego, CA). Significance is considered any value of p Յ 0.05.