Aminophospholipids are signal-transducing TREM2 ligands on apoptotic cells

Variants of triggering receptor expressed on myeloid cells 2 (TREM2) are associated with an increased incidence of Alzheimer’s disease, as well as other neurodegenerative disorders. Using a newly developed, highly sensitive reporter cell model, consisting of Jurkat T cells stably overexpressing a reporter gene and a gene encoding TREM2DAP12 fusion protein, we show here that TREM2-dependent signal transduction in response to apoptotic Neuro2a cells is mediated by aminophospholipid ligands, phosphatidylserine and phosphatidylethanolamine, which are not exposed on the intact cell surface, but become exposed upon apoptosis. We also show that signal-transducing TREM2 ligands different from aminophospholipids, which appear to be derived from neurons, might be present in membrane fractions of mouse cerebral cortex. These results may suggest that TREM2 regulates microglial function by transducing intracellular signals from aminophospholipids on apoptotic cells, as well as unidentified ligands in the membranes of the cerebral cortex.

in the cytosolic domain. TREM2 is therefore postulated to bind ligands at its extracellular immunoglobulin V-type domain and to transduce intracellular signals that regulate microglial functions such as cytokine production, migration, proliferation, phagocytosis, cell survival, synapse elimination, and compaction of amyloid plaques 23,[27][28][29][30][31][32][33][34] . Notably, inflammatory cytokines are downregulated in Trem2-knockout mice 30,31,33 , and anti-Trem2 agonistic antibody induces inflammatory cytokines in the brain 35 . These results suggest that TREM2 has a role in the production of inflammatory cytokines in vivo. Interestingly, a functional interaction between TREM2 and familial AD gene presenilin 1 has been reported 36 . Proper signal transduction might be disturbed by the risk-associated variants, leading to increased incidence of neurodegenerative disorders. Many proteins or compounds that bind to the TREM2 extracellular domain have been reported, including phosphatidylserine (PS), phosphatidylethanolamine (PE), phosphatidylcholine (PC), sulfoglycolipid, apolipoproteins, low-density lipoprotein, high-density lipoprotein, heat shock protein 60, DNA, E. coli, apoptotic cells, and Aβ 31,37-54 . Among them, apoptotic cells 31,40 , normal cultured cells 39,40 , and oligomeric Aβ 53,54 have been confirmed to induce TREM2-mediated intracellular signalling. However, the ligands expressed by apoptotic or normal cultured cells have not been identified, and it is not clear whether other signal-transducing ligands of TREM2 are present in the brain. In addition, further investigations are required to understand the different pathological effects of TREM2 variants on the signal transduction 50 .
In this work, we have developed a highly sensitive reporter cell model to monitor signal transduction from TREM2, in order to identify the TREM2 ligands on apoptotic and normal cultured cells and to search for other ligands in the brain that may be associated with increased incidence of AD. We found that the ligands expressed on apoptotic Neuro2a cells, as well as some normal cultured cells, are aminophospholipids. Finally we suggest that other TREM2 signal-transducing ligands are present in the membranes of mouse cerebral cortex.

Results
Establishment of reporter cell models for TREM2-dependent signal transduction. To establish sensitive and specific reporter cell models to monitor TREM2 signal transduction, RAW264.7, 2B4, and Jurkat cells were stably transfected with reporter genes: nuclear factor of activated T cells (NFAT) promoter element and TREM2DAP12 fusion cDNA corresponding to TREM2 extracellular domain and DAP12 transmembrane and cytosolic domain 38,55,56 . RAW264.7 was chosen because it is a macrophage cell line having similar characteristics to microglia, such as phagocytosis and cytokine production. 2B4 and Jurkat cell lines were chosen because they are derived from T cells, which are the only cell type so far shown to transduce signals via TREM2 31,37,40 . These reporter cells were incubated with apoptotic Neuro2a cells to induce signal transduction via TREM2 40 and luminescence or fluorescence of the cell lysates was measured. RAW264.7 reporter cells expressing TREM2DAP12 showed increased luminescence upon apoptotic cell treatment, but RAW264.7 cells not expressing TREM2DAP12 also showed increased luminescence, indicating that the signal is independent of exogenous TREM2DAP12 (not shown). On the other hand, 2B4 reporter cells expressing TREM2DAP12 showed a 2-to 3-fold increase of fluorescence upon apoptotic cell treatment, while the parent 2B4 cells (control) did not (not shown). More strikingly, Jurkat reporter cells expressing TREM2DAP12 showed a luminescence increase of more than 10-fold, while the parent cells did not (see Figs 1 to 3). These results confirmed that apoptotic cells express TREM2 ligand(s) that transduce an intracellular signal leading to NFAT activation 31,40 , and that T cell lines are suitable to study TREM2 signal transduction from apoptotic cells. In the following study, we used Jurkat reporter cells because of the high signal-to-background ratio.
We first investigated the time course of the NFAT activation by apoptotic cells. As shown in Fig. 1, luminescence of Jurkat reporter cell clones expressing TREM2DAP12 began to increase at 6 hours and reached a plateau at 14 hours when equal numbers of apoptotic cells and reporter cells were incubated. We then examined whether or not the signal is TREM2-dependent. Reporter cell clones expressing TREM2DAP12 or TREM1DAP12 or parent reporter cells ( Fig. 2A) were incubated together with apoptotic cells, and the luminescence of the cell lysates was measured. As shown in Fig. 2B, two independent reporter cell clones expressing TREM2DAP12 showed increased luminescence upon apoptotic cell treatment, while the parent reporter cells and two independent reporter cell clones expressing TREM1DAP12 did not, even though the amounts of TREM1DAP12 www.nature.com/scientificreports www.nature.com/scientificreports/ proteins expressed in the reporter cells were much greater than those of TREM2DAP12 proteins ( Fig. 2A). All of five independent TREM2DAP12-expressing cell clones that we established showed increased luminescence in response to the apoptotic cells (not shown), although the efficacy of signal transduction was different in each case. Moreover, anti-TREM2 antibody inhibited the reporter signal in a dose-dependent manner (Fig. 2C). These results indicate that the signal transduced in the TREM2DAP12-expressing reporter cells exposed to apoptotic cells is TREM2-dependent.
TREM2 ligands expressed on apoptotic cells are aminophospholipids. Next, we examined the characteristics of the TREM2 ligands expressed in apoptotic cells. We incubated the reporter cells with apoptotic or normal cultured Neuro2a cells and measured the intensity of luminescence in the cell lysates. We found that apoptotic Neuro2a cells transduced the signal in a dose-dependent manner, while normal cultured Neuro2a cells transduced a signal similar to or weaker signal than that obtained with medium alone (Fig. 3A). When the reporter cells and apoptotic cells were cultured without direct contact (separated by a culture insert membrane with 0.4 μm pore size), the signal was almost completely blocked (Fig. 3B). These results suggest that TREM2 ligands are exposed on the cell surface upon apoptosis, and are not factors secreted from the cells.
These characteristics of TREM2 ligands are reminiscent of aminophospholipids, PS or PE, which are localized in the inner leaflet of the plasma membrane in intact cells, but become exposed when the cells are triggered to undergo apoptosis 57 . Recombinant protein MFGE8-L D89E, which retains the binding domain to . TREM2 ligands are exposed at the cell surface upon apoptosis. (A) Apoptotic (closed circles) or normal (open circles) Neuro2a cells and reporter cell clones expressing TREM2DAP12 were incubated with the indicated ratio of Neuro2a: the reporter cell numbers and the luminescence of cell lysates were measured. Data are shown as means ± SD (n = 3). Where SDs are not shown, they are smaller than the symbols. (B) Apoptotic cells and reporter cell clones expressing TREM2DAP12 were incubated at a ratio of 1/10 with or without culture inserts (0.4 μm pore size) and the luminescence of cell lysates was measured. Data are shown as means ± SD (n = 4). One-way ANOVA with the Student-Newman-Keuls test was performed. *p < 0.05. . Two-tailed Student's t test was performed. (C) Apoptotic cells and Jurkat TREM2DAP12 reporter cell clones were incubated at a ratio of 1/10 together with anti-TREM2 antibodies and the luminescence of the cell lysates was measured. Data are shown as means ± SD (n = 3). One-way ANOVA with the Student-Newman-Keuls test was performed. ***p < 0.001. aminophospholipids but cannot bind phagocytes expressing integrins 58 , was purified to near homogeneity ( Fig. 4A) and added to the culture medium of the reporter cells and apoptotic cells. As shown in Fig. 4B, MFGE8-L D89E completely inhibited the signal induced by apoptotic cells, while control protein (BSA) did not. This strongly indicated that the TREM2 ligands on the apoptotic cells are indeed aminophospholipids. In accordance with this observation, pure aminophospholipids PS and PE induced signal transduction in the reporter cells, while PC, which is located on the exterior of intact cells, did not (Fig. 4C). The EC50 values for PS and PE could not be determined because the reporter cells died before the luminescence reached a plateau in the presence of higher concentrations of the lipids (not shown).
TREM2 ligands expressed on normal cultured cells are aminophospholipids. We then explored whether or not normal cultured cell lines express signal-transducing ligands of TREM2, because some cultured cell lines were shown to transduce NFAT signalling via TREM2 39,40 . We incubated the reporter cells with various cell lines and measured the luminescence. As shown in Fig. 5A normal cultured Jurkat and THP-1 cells induced signal transduction in TREM2DAP12-expressing reporter cells although the strength of luminescence was much less than that of apoptotic cells. This signal was TREM2-dependent, because TREM1DAP12-expressing reporter cells showed no signal transduction (Fig. 5B). When MFGE8-L D89E protein was added to the cultures, the signal was inhibited (Fig. 5C), suggesting that the TREM2 ligands on these normal cultured cells are also aminophospholipids, which might have become exposed during normal culture (see discussion).
TREM2 ligands expressed in the brain are not aminophospholipids. Finally, we searched for TREM2 signal-transducing ligands in the brain. We incubated homogenates of cerebral cortices from young adult mice (5 months old) with reporter cells expressing TREM2DAP12 and measured the luminescence. We found that homogenates prepared from three independent mice increased the reporter activity ( Fig. 6A) suggesting that the cerebral cortex contains ligand(s) that induce NFAT signalling via TREM2. This signal was TREM2-dependent, because reporter cells expressing TREM1DAP12 (Fig. 6B) or parent control cells (not shown) showed no signal. We then found that the ligands were present in the membrane fraction, not soluble fraction of the homogenates (Fig. 6C). Unexpectedly, the signal induced by the membrane fraction was not inhibited by MFGE8-L D89E (Fig. 6D); this result suggests that the ligands in the mouse cerebral cortical membranes are different from aminophospholipids. To examine if the ligands are expressed by astrocytes, microglia or neurons, we compared the TREM2 signal-inducing activity of cerebral cortices derived from wild-type and App knock-in mice carrying NL-G-F mutations (App NL-G-F ) 59 , since the App NL-G-F mice have increased numbers of astrocytes and microglia in an aging-dependent manner after 6 months 60 . However, we found no significant difference of TREM2 signal transduction between the two strains at 6 or 12 months (Fig. 6E) suggesting that the ligands might not be expressed by astrocytes or microglia.

Discussion
We have established a highly sensitive and specific reporter cell model for TREM2 signal transduction in Jurkat T cells. So far, two cell lines, BWZ and 2B4 T cells, have been reported as TREM2 signal-transducing cell models 31  Apoptotic cells and reporter cell clones expressing TREM2DAP12 were incubated at a ratio of 1/10 together with 1 μg/mL MFGE8-L D89E or BSA as a control, and the luminescence of the cell lysates was measured. Data are shown as means ± SD (n = 3). One-way ANOVA with the Student-Newman-Keuls test was performed. (C) The reporter cell clones expressing TREM2DAP12 were cultured in wells coated with purified PS, PE or PC and the luminescence of the cell lysates was measured. Data are shown as means ± SD (n = 3). One-way ANOVA with the Student-Newman-Keuls test was performed. ***p < 0.001.
www.nature.com/scientificreports www.nature.com/scientificreports/ established in this study is even more sensitive, as the luciferase activity was increased by more than 10-fold in the presence of apoptotic cells (Figs 1 to 3). Possible reasons for this are that Jurkat cells may have greater amounts of signalling molecules required for NFAT signal transduction via TREM2, or that TREM2DAP12 fusion proteins are better able to transduce the signal than independently expressed TREM2 and DAP12. Another possibility might be that the luc2P reporter gene used in this study has humanized codons, which may be favorable for higher expression and reduced anomalous transcription. The luc2P gene also contains hPEST, a protein destabilization sequence, which allows luc2P protein levels to respond more quickly. A further advantage is that enzymatic detection of the luc2P is faster (5 min) than that of LacZ 40 . In addition, the procedure is simpler than sorting EGFP-expressing cells 31 , and is suitable for high-throughput screening of the agonists/antagonists. At the beginning of this study we tried to establish reporter cell models in RAW264.7 macrophage cell lines, but these cells transduced the signal irrespective of overexpression of TREM2DAP12. Endogenous expression of TREM2 and DAP12 in RAW264.7 cells 61,62 might inhibit the signal from the overexpressed TREM2DAP12. As we and others have shown, T cell lines are suitable models for TREM2 signal transduction, probably because T cells express little endogenous TREM2 or DAP12 and share common signalling molecules with microglia, considering that T cell receptors and DAP12 have the same ITAM at the cytoplasmic domains.
The Jurkat reporter cells expressing TREM2DAP12 transduced the signal when incubated with apoptotic cells, while those expressing TREM1DAP12 did not (Fig. 2B), indicating the specificity of these reporter cells. We speculate that ligand recognition involves specific amino acid residue(s) in TREM2. Identification of the ligand-binding site in TREM2 would facilitate an understanding the interaction between TREM2 and the ligands, and would be helpful for drug development. The TREM2 antibody clone 237920 works as an antagonist of the NFAT signal via TREM2 induced by apoptotic cells (Fig. 2C). This may suggest that the epitope of the antibody is located near the binding site to apoptotic cells. Thus, mapping of the epitope might help to identify the ligand-binding site in TREM2. It would be interesting to see whether or not the epitope is identical to the reported PS-binding site in TREM2 extracellular domain 26 .
Apoptotic cells transduce the NFAT signal via TREM2 31,40 , while aminophospholipids such as PS and PE also transduce the signal via TREM2 31,48 . However, it has not been reported whether the TRME2 ligands in apoptotic cells are aminophospholipids. In this study we established for the first time that the TREM2 ligands in apoptotic cells are aminophospholipids by demonstrating inhibition of the signal with MFGE8 (Fig. 4B), an aminophospholipid-binding protein that inhibits the interaction between apoptotic cells and phagocytes 58 . This conclusion is supported by our findings that purified aminophospholipids (PS and PE), but not PC, transduce the signal (Fig. 4C), and that direct binding between reporter cells and apoptotic cells is required for signal transduction (Fig. 3B), because aminophospholipids are exposed on the cell surface after apoptosis, while PC is already exposed in healthy cells. Actually, Bailey reported that TREM2 binds to PS but not to PC 45 . However, other reports suggested that PC might induce signal transduction via TREM2 31,48 . The apparent contradiction may be due to the different reporter cell systems used (2B4 cells in their studies vs. Jurkat cells in this study). Bader Lange et al. reported that larger amounts of PS and PE are exposed on the surface of the synaptosomes in aged and AD model mouse brain 63 , and this may increase TREM2 signal transduction, leading to expression of microglial protective phenotypes. The increased incidence of AD associated with TREM2 R47H or R62H might be due to impaired protection of the TREM2 signal during aging or AD progression resulting from loss of function 50 . Although we still do not know exactly the target genes of TREM2 signal transduction responsible for the increased risk of AD, www.nature.com/scientificreports www.nature.com/scientificreports/ genes encoding factors involved in mTOR activation, which controls energetic and anabolic metabolism, might be candidates 64 . Further study will be needed to understand the link between TREM2 signal transduction and AD pathogenesis at the molecular level.
Our results indicate that the TREM2 ligands on normal cultured Jurkat and THP-1 cells are aminophospholipids (Fig. 5C). We hypothesize that small amounts of cells might die spontaneously during normal culture, resulting in exposure of previously inaccessible aminophospholipids. Since TREM2DAP12-expressing reporter cells are derived from Jurkat cells, aminophospholipids exposed on small amounts of the dying reporter cells might transduce the NFAT signal by themselves. Indeed, MFGE8 reduced the reporter activity to less than that of the normal cultured reporter cells (compare 3 rd and 5 th columns with 1 st column in Fig. 5C), supporting this hypothesis. Unknown TREM2 ligands on normal cultured cells that transduce the NFAT signal were previously reported 39,40 . Our results suggest that those unidentified ligands might be aminophospholipids.
Finally, we showed that TREM2 ligands that can specifically transduce intracellular signalling are present in the membrane fractions of 5-month-old mouse cerebral cortex, and that MFGE8 protein does not inhibit the signal (Fig. 6), suggesting that the signal-transducing ligands in the mouse brain are different from aminophospholipids. However, it is not clear why aminophospholipids, which should be present in the membrane fractions, cannot transduce the signal. One possible explanation would be that signal-transducing ligands in the brain may have a higher affinity for TREM2 than aminophospholipids, thus excluding the signal from aminophospholipids. Since nuclear and soluble fractions were removed from the membrane fractions, DNA and RNA can be excluded as candidates for the endogenous signal-transducing ligands. Subtle quantitative or qualitative alterations of the ligands might occur in the brain during aging, resulting in an increase in the incidence of AD. The TREM2 signal-inducing activities in the cerebral cortex from wild-type and App NL-G-F knock-in mice were not significantly different (Fig. 6E), suggesting that the ligands might not be derived from astrocytes or from microglia, considering that the numbers of astrocytes and microglia are robustly increased after 6 months in App NL-G-F knock-in mice 60 . Thus, the ligands might be expressed by live neurons. Identification of the ligands in the cortices will be an important next step for understanding the functional interaction between microglia and neurons, as well as the relevance of TREM2 to the development of neurodegenerative disorders. It would also be interesting to investigate whether AD-related variants such as R47H affect the binding and/or signal transduction from the ligands in the mouse cerebral cortex, and we are currently comparing the signal transduction characteristics of reporter cell clones expressing wild-type TREM2DAP12 and R47H variant.
In sum, we have developed a highly sensitive reporter cell model for signal transduction via TREM2 and showed that the TREM2 ligands on apoptotic and normal cells are aminophospholipids. The model should be useful to examine signal transduction by apolipoprotein E or oligomeric Aβ and to investigate the effects of TREM2 variants on the signal transduction. www.nature.com/scientificreports www.nature.com/scientificreports/ Methods cDNAs. Human TREM2 65 and mouse MFGE8-L D89E 58 cDNAs were as described. Human TREM1 and DAP12 cDNAs were obtained from Kazusa DNA Research Institute (Chiba, Japan) and Health Science Research Resources Bank (Osaka, Japan), respectively. cDNAs encoding the extracellular domain of TREM1 (amino acid residue 1-203) and TREM2 (amino acid residue 1-172) were amplified, fused with the transmembrane and cytosolic domain of human DAP12 (amino acid residues 30-113) with its stop codon by PCR, and inserted into the NheI and XhoI sites of pEBMulti-Neo TARGET tag-C (Fuji Film, Osaka, Japan). pGL4.30[luc2P/NFAT-RE/ Hygro] vector (Promega, Madison, WI, USA) was used as a reporter gene, containing the luciferase reporter gene luc2P (Photinus pyralis) under the control of NFAT-responsive element.
2B4 cells stably transfected with enhanced green fluorescent protein (EGFP) gene under the control of NFAT-responsive element 66 were transfected with TREM2DAP12 cDNA using a NEPA21 super electroporator and selected with 1 mg/mL G418.
Reporter assay. Neuro2a cells were treated with 100 nM staurosporin (Fuji Film) overnight to induce apoptosis, then washed and resuspended in the culture media. The apoptotic Neuro2a cells and 1 × 10 5 Jurkat reporter cells were incubated at 37 °C in a ratio of 1/16 to 2/1 in 48-well plates for 14 hours, unless otherwise indicated. The cell lysates were incubated with One-Glo TM reagent (Promega) in a 384-well plate for 5 min, and the luminescence was measured with a SPARK 20 M (Tecan, Männedorf, Switzerland). Normal cultured Neuro2a, Jurkat, THP-1, RAW264.7, and HEK293 cells were collected by pipetting or with TrypLE ™ Select CTS ™ (Thermo Fisher Scientific) and resuspended in the culture media after centrifugation. Where indicated, anti-TREM2 antibody (clone 237920, R&D Systems, Minneapolis, MN, USA), purified milk fat globule-EGF factor 8 (MFGE8)-L D89E, or bovine serum albumin (BSA) (Takara Bio, Osaka, Japan) was added to the culture media. Culture inserts with 0.4 μm pore membrane (Thermo Fisher Scientific) were used to block direct interaction between the apoptotic cells and reporter cells. Phosphatidylserine (PS), phosphatidylethanolamine (PE) and phosphatidylcholine (PC) (Sigma-Aldrich) were dissolved in methanol and coated on plates by drying before the reporter cells were seeded. Homogenates of cerebral cortices from wild-type and App NL-G-F knock-in 59 mice were prepared by homogenization in buffer (50 mM Tris-HCl buffer (pH 7.4), 150 mM NaCl, protease inhibitor cocktail, 0.7 ug/ml pepstatin A). The homogenates were centrifuged at 800 g for 10 min at 4 °C to remove nuclei and cell debris, and the post nuclear supernatants were centrifuged at 70,000 rpm in a TLA110 rotor (Beckman Coulter Diagnosis, Brea, CA, USA) for 29 min at 4 °C. The supernatants (soluble proteins) and resuspended membrane pellets were incubated with the reporter cells. All experiments conducted here were approved by the Ethics Committees of Nagasaki University.
Preparation of MFGE8-L D89E protein. HEK293 cells were transfected with MFGE8-L D89E cDNA 58 using Lipofectamine3000 (Thermo Fisher Scientific). After 24 hours, the medium was replaced with DMEM without fetal bovine serum and the cells were further cultured for 30 hours. MFGE8-L D89E protein in the culture media was purified with anti-DDDDK-tag mAb-Magnetic Beads (MBL, Aichi, Japan) and subjected to silver staining (Silver Stain MS kit (Fuji Film). Statistical analysis. Statistical analysis was done by SigmaPlot software, ver.14.0 (Systat Software Inc., San Jose, CA). To compare two groups, a two-tailed Student's t test was employed after confirming equality of variances of the groups. To compare more than two groups, we used one-way ANOVA with the Student-Newman-Keuls test. P values of less than 0.05 were considered to be significant.