Research ArticleNeurodegenerative Disease

TREM2 mutations implicated in neurodegeneration impair cell surface transport and phagocytosis

See allHide authors and affiliations

Science Translational Medicine  02 Jul 2014:
Vol. 6, Issue 243, pp. 243ra86
DOI: 10.1126/scitranslmed.3009093


Genetic variants in the triggering receptor expressed on myeloid cells 2 (TREM2) have been linked to Nasu-Hakola disease, Alzheimer’s disease (AD), Parkinson’s disease, amyotrophic lateral sclerosis, frontotemporal dementia (FTD), and FTD-like syndrome without bone involvement. TREM2 is an innate immune receptor preferentially expressed by microglia and is involved in inflammation and phagocytosis. Whether and how TREM2 missense mutations affect TREM2 function is unclear. We report that missense mutations associated with FTD and FTD-like syndrome reduce TREM2 maturation, abolish shedding by ADAM proteases, and impair the phagocytic activity of TREM2-expressing cells. As a consequence of reduced shedding, TREM2 is virtually absent in the cerebrospinal fluid (CSF) and plasma of a patient with FTD-like syndrome. A decrease in soluble TREM2 was also observed in the CSF of patients with AD and FTD, further suggesting that reduced TREM2 function may contribute to increased risk for two neurodegenerative disorders.


Homozygous loss-of-function mutations in the triggering receptor expressed on myeloid cells 2 (TREM2) gene cause polycystic lipomembranous osteodysplasia with sclerosing leukoencephalopathy [also known as Nasu-Hakola disease (NHD)], a disease characterized by ankle swellings and frequent bone fractures (1). During disease progression, NHD patients develop neurological syndromes reminiscent of the behavioral variant of frontotemporal dementia (FTD) (1). Recently, homozygous missense mutations of TREM2, such as the p.T66M and the p.Y38C mutations, as well as a compound heterozygous missense mutation have been identified to cause an FTD-like syndrome without bone pathology (2, 3). Genetic screenings have now also identified heterozygous missense mutations in TREM2 as risk factors for Alzheimer’s disease (AD), Parkinson’s disease (PD), amyotrophic lateral sclerosis (ALS), and FTD (49). Most reported missense mutations are located in the ectodomain of TREM2, a membrane-bound type 1 protein (Fig. 1A). Integrated network analysis suggested a central role for TREM2 in various brain areas (10), where it is mainly expressed in microglia cells regulating essential functions including phagocytosis and the removal of apoptotic neurons (1114).

Fig. 1. TREM2 is shed by ADAM10 in HEK293 Flp-In cells.

(A) Illustration of membrane-bound TREM2. Upon shedding by ADAM10, the remaining C-terminal stub of TREM2 is cleaved within the membrane by γ-secretase (17). Ig, immunoglobulin; HA, hemagglutinin; TM, transmembrane domain; ICD, intracellular domain. (B) Pharmacological inhibition of ADAM proteases using a broad ADAM inhibitor (GM 6001) or a more selective ADAM10 inhibitor (GI 254023X) but not a more selective ADAM17 (GL 506-3) or BACE1 (C1) inhibitor reduced sTREM2 generation and stabilized fully glycosylated, cell surface–associated, mature membrane-bound wild-type (WT) TREM2. (C) Small interfering RNA (siRNA)–mediated ADAM10 knockdown confirmed reduced ADAM10-mediated TREM2 shedding. Hashtags indicate nonspecific immunoreactivity. Anti-calnexin antibody (B and C) was used as a loading control. Bar graphs (B and C) show enzyme-linked immunosorbent assay (ELISA) quantification of sTREM2 relative to control-treated cells. Quantitative data are represented as means ± SD from at least two independent experiments; n = 4 to 7 (B) and n = 8 (C). Statistical differences were calculated by Mann-Whitney U test. **P < 0.01.

The TREM2 homolog TREM1 is found as a soluble variant in the serum of patients with septic shock and is secreted from monocytes upon stimulation with lipopolysaccharide (15). Furthermore, TREM2 was also observed in plasma and cerebrospinal fluid (CSF) samples from patients with multiple sclerosis (16). In line with these findings, recent evidence suggests that TREM2 is a substrate for regulated intramembrane proteolysis (RIP) (17). RIP substrates are membrane-bound proteins, whose ectodomains are released upon shedding by proteases such as members of the ADAM (a disintegrin and metalloproteinase domain-containing protein) or BACE (β-site APP cleaving enzyme) family (18). Upon removal of the ectodomain, the remaining membrane-retained stub is further processed by intramembrane proteolysis (19).

Whether and how missense mutations affect TREM2 function is elusive. Here, we investigated whether TREM2 missense mutations found in patients with FTD, FTD-like syndrome without bone involvement, AD, and other neurodegenerative diseases affect the transport and processing of the TREM2 protein and thus may cause its loss of function.


Generation of soluble TREM2 by ADAM10-mediated ectodomain shedding

Soluble fragments of TREM2 (sTREM2) have been observed in supernatants of dendritic cell cultures as well as in plasma and CSF samples from patients with noninflammatory neurological diseases and multiple sclerosis (16). Consistent with that, we observed secreted fragments of TREM2 (Fig. 1B) migrating as multiple bands in the range of 36 to 50 kD as well as the full-length membrane-bound TREM2 migrating between 36 and 60 kD (Fig. 1B) in isogenic human embryonic kidney 293 cells (HEK293 Flp-In) stably expressing a single copy of human wild-type TREM2. We did not find C-terminal FLAG epitope containing fragments in the conditioned medium, implying that sTREM2 is produced by ectodomain shedding (fig. S1). Moreover, in HEK293 Flp-In cells, the broad ADAM inhibitor GM 6001 reduced secretion of sTREM2 as did the more selective ADAM10 inhibitor GI 254023X, but not the ADAM17-specific inhibitor GL 506-3 or the BACE inhibitor C3 (Fig. 1B). Reduced sTREM2 correlated with an increase in fully glycosylated mature membrane-bound TREM2 (Fig. 1B, middle panel; for further analysis of immature and mature TREM2, see the pulse-chase experiment in Fig. 2E and the deglycosylation experiments in fig. S2). siRNA-mediated knockdown revealed ADAM10 as a major sheddase of TREM2 in HEK293 Flp-In cells (Fig. 1C).

Fig. 2. Reduced cell surface transport and shedding of mutant TREM2.

(A) Anti-TREM2 and anti-HA immunoblotting of sTREM2 in supernatants from cells expressing the FTD- and FTD-like–associated TREM2 mutations p.T66M and p.Y38C or mutations of conserved cysteine residues at positions 36 and 60. Lower panel shows quantification of sTREM2 by ELISA on supernatants from stable HEK293 Flp-In cells. (B) Immature mutant TREM2 (black arrowhead) accumulated, whereas the mature form and CTFs were reduced compared to WT TREM2. An alternative proteolytic event appears to produce minor amounts of a larger CTF (gray arrowhead). (C) Quantification of immunoblots shows a small increase in immature p.R47H TREM2 and a significant decrease in the ratio of mature-to-immature TREM2. (D) Coexpression of DAP12, the signaling adaptor of TREM2, did not affect reduced maturation and secretion of mutant TREM2. (E) Pulse-chase experiments revealed a longer half-life for mutant TREM2 within cell lysates accompanied by a minimal release of sTREM2 into the medium. (F) Surface biotinylation of mature surface-exposed mutant TREM2. Anti-APP (A) or anti-calnexin (B) antibodies were used as loading controls for supernatants or membrane fractions, respectively. Anti-calnexin antibody was used to prove selective cell surface labeling. NT, nontransfected HEK293 Flp-In host cell line. Quantitative data are represented as means ± SD from three independent experiments using three independent cell lines for WT or p.R47H TREM2. n > 6 (A) and n = 9 (C). Statistical differences were calculated by Mann-Whitney U test. **P < 0.01; ***P < 0.001.

Impaired cell surface transport of mutant TREM2

Expression of the FTD and FTD-like TREM2 p.T66M and p.Y38C mutations (2, 9) revealed a strong reduction of sTREM2 in conditioned medium from isogenic HEK293 Flp-In cells (Fig. 2A). Furthermore, the TREM2 p.R47H mutation, which increases the risk for AD, PD, and ALS (47), also showed reduced secretion of sTREM2 albeit to a lower extent (Fig. 2A). In parallel, we observed an accumulation of membrane-bound immature full-length TREM2 p.T66M and p.Y38C together with a reduction in TREM2 C-terminal fragments (CTFs) generated by ADAM10-mediated proteolytic cleavage (Fig. 2B). The detection of accumulating amounts of immature mutant TREM2 variants also confirmed that the mutations do not abolish antibody recognition. Consistent with a weaker effect of the p.R47H mutation on TREM2 shedding, we observed a much less marked accumulation of immature p.R47H mutant TREM2 (Fig. 2B; quantitated in Fig. 2C). Mutagenesis of conserved cysteine residues (p.C36A and p.C60A) blocked TREM2 shedding like the p.T66M and p.Y38C mutations (Fig. 2A) and caused an accumulation of immature membrane-bound full-length TREM2 as well as a reduction of the CTF (Fig. 2B). This further supports the possibility that mutant TREM2 is misfolded and retained within the cell. Coexpression of DAP12, the signaling adaptor for TREM2 (20), which is not expressed in HEK293 Flp-In cells, did not restore transport and shedding of mutant TREM2 (Fig. 2D). Pulse-chase experiments showed that wild-type membrane-bound full-length TREM2 matured from a low–molecular weight immature species to a higher–molecular weight mature species most likely by glycosylation (Fig. 2E and fig. S2). Upon maturation after 60 to 90 min of cold chase, TREM2 was shed and accumulated over time in the conditioned medium (Fig. 2E). In contrast to wild-type TREM2, mutant p.T66M TREM2 failed to mature efficiently, resulting in the accumulation of immature TREM2 in the cell lysate and inefficient generation of sTREM2 (Fig. 2E). Accordingly, cell surface TREM2 was decreased upon expression of the TREM2 p.T66M, p.Y38C, p.C36A, and p.C60A mutants compared to wild-type TREM2 (Fig. 2F). Consistent with less marked effects of the p.R47H mutation on TREM2 maturation and sTREM2 generation, p.R47H TREM2 showed similar cell surface expression compared to wild-type TREM2 (Fig. 2F). Immunohistochemistry also showed severely reduced cell surface expression of TREM2 p.T66M, p.Y38C, p.C36A, and p.C60A accompanied by an increase of intracellular staining, which colocalized preferentially with the endoplasmic reticulum (ER) marker calnexin (Fig. 3A). Consistent with our biochemical findings, the p.R47H TREM2 mutation showed less intracellular accumulation (Fig. 3A), and consequently, robust expression of cell surface TREM2 was detected (Fig. 3B). Thus, mutations associated with FTD and FTD-like syndrome affected TREM2 maturation, cell surface transport, and proteolytic processing, whereas the AD-associated p.R47H mutation had a mild effect on maturation and secretion, which is at the limit of biochemical detection.

Fig. 3. Mutant TREM2 accumulates in the cytoplasm of stable HEK293 Flp-In cells.

(A) Double immunofluorescence of TREM2 labeled with anti-HA antibody and anti-calnexin antibody. White boxes on the merged images depict the enlarged area shown in the images to the right. White arrows in enlarged images highlight selected areas showing TREM2 and calnexin colocalization. (B) Surface staining of WT and mutant TREM2-expressing HEK293 Flp-In cells. Scale bars, 10 μm. DAPI, 4′,6-diamidino-2-phenylindole.

Reduced maturation and shedding of mutant TREM2 in microglial cells

Within the human brain, TREM2 is primarily expressed in microglia cells (1114). Therefore, we aimed to confirm our findings in an in vivo relevant setting. To do so, we expressed wild-type TREM2, as well as the TREM2 mutations p.T66M, p.Y38C, p.R47H, and p.C36A, in the murine microglial BV2 cell line, a model frequently used for in vitro studies of microglial function (2123). Reverse transcription polymerase chain reaction (RT-PCR) analysis confirmed the expression of selected microglial markers CD11b and CD68 as well as the expression of Trem2 and Dap12 in the BV2 cell line (Fig. 4A). The p.T66M, p.Y38C, and p.C36A mutations exhibited reduced shedding of sTREM2 in BV2 cells (Fig. 4B), as observed in HEK293 Flp-In cells (Fig. 2A). Remarkably, the TREM2 p.R47H mutant consistently showed reduced expression in the membrane fraction of transiently transfected BV2 cells (Fig. 4B), an observation that was also confirmed in stably transfected BV2 cell lines (fig. S3). Nevertheless, similar to our findings in HEK293 Flp-In cells (Fig. 2A), the p.R47H mutation showed reduced secretion of sTREM2 (Fig. 4B). Shedding of TREM2 by microglia cells was also mediated by proteases of the ADAM family but not BACE1 (Fig. 4C). In the microglia BV2 cell line, both ADAM10 and ADAM17 contributed to shedding of TREM2 (Fig. 4C). We also confirmed reduced cell surface transport of mutant TREM2 in BV2 cells by investigating wild-type and p.T66M mutant TREM2 in a cell surface biotinylation assay. In line with our findings in HEK293 Flp-In cells, we found less cell surface TREM2 upon expression of the p.T66M mutation (Fig. 4D). Moreover, immunohistochemistry also fully confirmed the retention of mutant TREM2 in BV2 cells (Fig. 4E). Although wild-type TREM2 was predominantly observed in the Golgi, the p.T66M, p.Y38C, and p.C36A mutant variants were retained predominantly within the ER (Fig. 4E). Consistent with reduced effects of the p.R47H mutation on maturation of TREM2 and sTREM2 generation, this variant was predominantly located within the Golgi (Fig. 4E). Thus, the effects of TREM2 mutations on reduced maturation and cell surface transport as well as on proteolytic processing were confirmed in the microglial BV2 cell line.

Fig. 4. Altered localization and reduced shedding of mutant TREM2 in BV2 microglial cells.

(A) RT-PCR analysis of Trem2, Dap12, and microglial markers (CD11b and CD68) in BV2 cells. (B) Comparison of TREM2 processing in BV2 cells transiently expressing WT or mutant TREM2. Note that expression of the p.R47H variant was lower in all conducted experiments. Quantification of sTREM2 by ELISA normalized to the expression of immature TREM2 (lower panel). The AD variant p.R47H also showed reduced sTREM2 generation, but to a lesser extent than the other studied mutants. Hashtag indicates a protein that cross-reacts with the human anti-TREM2 antibody. (C) Similar to the case with HEK293 Flp-In cells (see Fig. 1), pharmacologic inhibition of ADAM proteases with GM 6001 (broad ADAM inhibitor), GI 254023X (ADAM10 inhibitor), and GL 506-3 (ADAM17 inhibitor) reduced the generation of sTREM2 in the microglial BV2 cell line. Anti-calnexin antibody was used as a loading control for membrane fractions in (B) and (C). (D) Surface biotinylation of mature surface-exposed mutant TREM2. (E) Double immunofluorescence of TREM2 labeled with anti-HA antibody and anti-calnexin antibody. White boxes on the merged images depict the area enlarged in the images to the right. White arrows in enlarged images highlight selected areas with TREM2 and calnexin colocalization. Scale bars, 20 μm. Quantitative data are represented as means ± SD from at least two independent experiments; n = 4. Statistical differences were calculated by Mann-Whitney U test. *P < 0.05; n.s., nonsignificant.

Reduced sTREM2 in the CSF of patients with AD and FTD

To obtain evidence for reduced cell surface transport of mutant TREM2 in patients with FTD-like syndrome without bone involvement, we analyzed sTREM2 concentrations in the CSF and plasma of a patient carrying the TREM2 p.T66M mutation (2) as well as another patient with the p.Q33X mutation (2) for which only plasma was available. We established a highly sensitive sTREM2 ELISA (fig. S4, A to D) that showed good correlation with a previously published sTREM2 ELISA (16), which used independent antibodies (fig. S4B; Spearman rho = +0.521, P < 0.001). Consistent with the tissue culture analysis, these two independent ELISAs as well as immunoblotting revealed the virtual absence of sTREM2 in CSF from a patient with a homozygous TREM2 p.T66M mutation (Fig. 5A and fig. S4, E and F). Furthermore, the plasma concentration of sTREM2 in this patient was also below the detection limit and that of the patient carrying the p.Q33X mutation was very low (Fig. 5B). Because TREM2 is genetically linked not only to NHD and FTD-like syndrome without bone involvement but also to AD (4, 5) and FTD (6, 8, 9), we analyzed sTREM2 concentrations in CSF samples of a set of well-characterized FTD and AD patients (Table 1) and compared them with those of neurologically normal controls. Although we observed an overlap between both groups, statistical analysis revealed a significant reduction of sTREM2 in AD and FTD patients compared to control individuals (Table 1 and Fig. 5A; P = 0.001 and P < 0.001, respectively, Mann-Whitney U test). The significant decrease in sTREM2 concentrations in CSF in AD and FTD patients compared to controls was still present after controlling for the effect of gender and age, and was independent of the clinics collecting CSF [P = 0.001 and P < 0.001, respectively, analysis of covariance (ANCOVA)]. In contrast to CSF measurements, we did not detect any difference in sTREM2 concentrations in plasma among the controls and the AD and FTD patients (Table 2 and Fig. 5B; P = 0.872, Kruskal-Wallis).

Fig. 5. Reduced sTREM2 in the CSF of patients with FTD and FTD-like syndrome.

(A) ELISA-based analysis of sTREM2 in CSF samples shows virtual absence of sTREM2 in the TREM2 p.T66M mutation carrier, whereas robust concentrations of sTREM2 were detected in all control samples (n = 88). A reduction in sTREM2 was observed in AD patients (n = 56; Pcontrol vs. AD = 0.001) and FTD patients (n = 50; Pcontrol vs. FTD < 0.001). Horizontal bars indicate median sTREM2 concentrations per group with the interquartile range (IQR). (B) sTREM2 concentrations in plasma (nControl = 86; nAD = 51; nFTD = 35) showed no significant difference among these groups. Plasma from the homozygous TREM2 p.T66M mutation carrier was the only sample (1 of 174 samples) with undetectable sTREM2. Additionally, plasma from a homozygous TREM2 p.Q33X mutation carrier (2) also showed one of the lowest sTREM2 concentrations of all measured samples.

Table 1. Characteristics of patient and control study population used to measure CSF sTREM2.

Data are expressed as number of patients (percent), mean ± SD, or median (IQR) as appropriate. Probability values (P) denote differences between control, AD, and FTD patient groups. χ2 tests were used for gender. One-way analysis of variance (ANOVA) was used to compare age between groups followed by Tukey post hoc test. CSF biomarkers and sTREM2 were evaluated by nonparametric statistical analysis (Kruskal-Wallis and post hoc with Mann-Whitney U test).

View this table:
Table 2. Characteristics of patient and control study population used to measure plasma sTREM2.

Data are expressed as number of patients (percent), mean ± SD, or median (IQR) as appropriate. Probability values (P) denote differences between control, AD, and FTD patient groups. χ2 tests were used for gender. One-way ANOVA was used to compare age between groups followed by Tukey post hoc test. CSF biomarkers and sTREM2 were evaluated by nonparametric statistical analysis (Kruskal-Wallis and post hoc with Mann-Whitney U test).

View this table:

Impaired phagocytosis in TREM2 mutant cells

Reduction of cell surface TREM2 in microglia cells, the cell type in which TREM2 is selectively expressed within the brain, was reported to reduce removal of cellular debris and apoptotic neurons (14). In line with this, primary microglia from Trem2 knockout mice (Trem2−/−) (24), which do not produce detectable sTrem2 in conditioned medium (Fig. 6A), showed a reduced phagocytic capacity compared to wild-type controls in an assay using Escherichia coli conjugated to pHrodo that only yields a fluorescent signal in an acidic compartment (Fig. 6, B and C). Although the overall capacity of Trem2−/− microglia to phagocytose E. coli was only slightly reduced, Trem2−/− microglia seemed to be less competent in phagocytosing a larger amount of bacteria (Fig. 6, B and C).

Fig. 6. Impaired phagocytosis in microglia expressing mutant TREM2.

(A) ELISA-based analysis of sTrem2 in the supernatants of WT or Trem2 knockout (Trem2−/−) primary microglia. (B and C) Flow cytometric analysis of the phagocytic capacity of primary microglia using pHrodo E. coli as target particles. Q2, 50th percentile; Q3, 75th percentile. (D) Illustration of TREM2-DAP12 fusion construct (25) used in the phagocytosis assays shown in (F), (G), and (I). ITAM, immunoreceptor tyrosine-based activation motif. (E) Anti-TREM2 immunoblot of TREM2-DAP12–expressing HEK293 Flp-In cells. Anti-calnexin antibody was used as a loading control. (F) Phagocytosis of fluorescently labeled latex beads in HEK293 Flp-In cells stably expressing TREM2-DAP12 fusion constructs. Phagocytic index (percentage of cells that phagocytose beads expressed relative to WT) from three independent experiments is shown as mean ± SD. (G) Phagocytosis of pHrodo E. coli in HEK293 Flp-In cells stably expressing TREM2-DAP12 fusion constructs and quantified by flow cytometry. Data are depicted as means ± SD from at least two independent experiments. (H to J) Phagocytosis of 6-carboxyfluorescein (FAM)–labeled Aβ1–42 by (H) BV2 cells stably expressing WT or mutant TREM2, (I) HEK293 Flp-In cells stably expressing TREM2-DAP12 fusion constructs, and (J) primary microglia derived from Trem2−/− mice is shown as mean ± SD from two independent experiments (J) or three independent experiments (H and I) and is expressed relative to WT controls. In all assays, cytochalasin D (10 μM) was used as a negative control to inhibit phagocytosis. Statistical differences were calculated by Mann-Whitney U test. *P < 0.05; **P < 0.01; ***P ≤ 0.001.

Because our analysis in both the BV2 microglial cell line and the HEK293 Flp-In cells (Figs. 2 to 4) revealed reduced transport of mutant TREM2 to the cell surface, we next investigated whether this would correlate with a reduced capacity for phagocytosis. Consistent with that, bead uptake assays using HEK293 Flp-In cells expressing the ectodomain of TREM2 fused to the transmembrane and signaling domain of DAP12 (25, 26) (Fig. 6, D and E) demonstrated that reduced cell surface localization and secretion of mutant TREM2 (p.T66M and p.Y38C) directly correlated with a reduced phagocytic capacity using opsonized fluorescent latex beads (Fig. 6F; **P = 0.004, Mann-Whitney U test). Moreover, an independent assay using E. coli conjugated to pHrodo produced similar results (Fig. 6G and fig. S5). In line with our biochemical data in BV2 microglial cells as well as in HEK293 Flp-In cells, the AD risk variant p.R47H impaired phagocytosis to a much lesser extent in the bead uptake assay (Fig. 6F; **P = 0.004, Mann-Whitney U test) and was even normal for uptake of bacteria (Fig. 6G; n.s., P = 0.191, Mann-Whitney U test). Finally, we used the pathologically relevant amyloid β-peptide 1–42 (Aβ1–42) as a substrate in the phagocytosis assays and demonstrated that TREM2 loss of function due to the missense mutations significantly impaired phagocytosis of Aβ1–42 in three independent cell lines including BV2 microglial cells (Fig. 6H; ***P < 0.001, Mann-Whitney U test), HEK293 Flp-In cells (Fig. 6I; **P < 0.01, ***P < 0.001, Mann-Whitney U test), and Trem2−/− primary microglia (Fig. 6J; *P = 0.016, **P < 0.01, Mann-Whitney U test). Thus, mutations that severely reduce maturation of TREM2, such as p.T66M and p.Y38C, result in a significant loss of function of TREM2 as measured by three independent phagocytosis assays.

These findings suggest that expression of mature membrane-bound TREM2 correlates with phagocytic activity of microglia. Therefore, we reasoned that TREM2 may be modulated to increase clearance of cellular debris and amyloidogenic seeds. As proof of principle and as an additional link between cell surface TREM2 expression and phagocytosis, we investigated whether inhibition of TREM2 shedding, which increases cell surface TREM2 (Fig. 1, B and C), increases phagocytosis. In line with the data in Fig. 1, upon treatment of BV2 microglial cells endogenously expressing Trem2 (Fig. 4A), with a broad pharmacological inhibitor of ADAM proteases (GM 6001), proteolytic processing of endogenous Trem2 was significantly reduced (Fig. 7A; *P = 0.019; n.s., P > 0.05, Mann-Whitney U test). Moreover, reduced Trem2 shedding correlated with a significant increase in phagocytosis as shown by increased uptake of E. coli (Fig. 7B; ***P < 0.001; n.s., P > 0.05, Mann-Whitney U test). Thus, these findings provide further evidence for a link between cell surface expression of Trem2 and phagocytosis.

Fig. 7. Pharmacological inhibition of ADAM proteases decreased endogenous sTrem2 and increased phagocytosis in BV2 cells.

(A) Endogenous sTrem2 in supernatants from BV2 cells after inhibition of ADAM proteases (GM 6001) or BACE1 (C3). (B) Phagocytosis of pHrodo E. coli in BV2 cells treated with GM 6001 or C3. Data are represented as means ± SD from at least two independent experiments and expressed relative to nontreated control; n = 4 (A) and n = 7 to 9 (B). Statistical differences were calculated by Mann-Whitney U test. *P < 0.05; ***P < 0.001.


Loss of function of TREM2 is associated with NHD—a rare recessive disorder characterized by early-onset dementia with clinical presentation similar to the behavioral variant of FTD (1). Recently, two independent genome-wide association studies linked missense variants in TREM2 to an increased risk for developing late-onset AD (4, 5). Subsequent studies confirmed the association in several AD cohorts (27, 28) and further extended the finding to other neurodegenerative disorders including FTD (6, 8, 9), PD (6), and ALS (7). Although TREM2 variants are rare (population frequency ~0.3%), the effect size (odds ratio >3) is similar to that for APOEε4; in silico predictions suggest a probable damaging effect of the identified variants on TREM2 protein function (4, 8). Biochemical and cell biological analyses of two FTD-associated TREM2 variants (p.Y38C and p.T66M) indicated misfolding of TREM2 followed by inhibition of cell surface transport (Figs. 2 to 4). Reduced surface exposure of TREM2 leads to impaired phagocytosis (Fig. 6), a finding that is consistent with a short hairpin RNA–mediated TREM2 knockdown phenotype (14). Whereas the FTD- and FTD-like–associated mutations, as well as the mutations of the conserved cysteine residues, displayed a very severe and consistent biochemical and functional phenotype, the AD-associated p.R47H mutation showed a much weaker effect on maturation, secretion, and phagocytic activity. Whether this is due to a weaker general effect of this TREM2 variant or rather a different cellular mechanism of action remains to be investigated. Notably, the p.R47H mutation is considered to be a risk factor for AD (4, 5), whereas the other investigated variants are causative mutations when in the homozygous state (2). Furthermore, we observed that the TREM2 p.R47H mutant is expressed at lower rates, suggesting that this variant may be unstable and prone to degradation.

Our data suggest that reduced TREM2 function impairs phagocytosis, which may contribute to neurodegeneration through different mechanisms. First, reduced phagocytosis may prevent engulfment of cellular debris, which could result in a chronic inflammatory response (12). Second, amyloidogenic seeds, which are thought to prime neurodegeneration (29), may also be eliminated in a TREM2-dependent manner. Third, in the case of AD, TREM2 is up-regulated around amyloid plaques (4, 30), and it is tempting to speculate that this reflects a defense mechanism to remove unwanted protein aggregates. In line with this hypothesis, phagocytosis assays using preaggregated Aβ1–42 as a ligand suggest that TREM2 is indeed capable of removing amyloidogenic protein aggregates (Fig. 6, H to J). Finally, ER retention of mutant TREM2 may also cause ER stress, which could affect function and survival of microglia specifically in the case of the p.T66M and p.Y38C mutations.

TREM2 is a type I transmembrane glycoprotein that has been shown to shuttle to and from the plasma membrane in microglial cells upon cell stimulation by ionomycin or interferon-γ (31). Cell surface expression of TREM2 can be regulated by either phagocytic receptor recycling (23) or ectodomain shedding (Figs. 1, 4, and 7) during which sTREM2 fragments are released from the cell. Whether sTREM2 has a paracrine signaling function or serves as a competitor for TREM2 ligands, as is the case for TREM1 (12), remains unclear. However, such functions would also be compromised by the reduced secretion of TREM2 variants investigated here. sTREM2 can be readily detected by ELISA-based methods (16) and could therefore serve as a possible marker for neurodegenerative disorders in the future, although further validation in additional patient cohorts is needed. A patient with an FTD-like syndrome associated with the p.T66M mutation showed no sTREM2 in either CSF or plasma. Consistent with a loss of function of TREM2, homozygous DAP12 deletions are also found in patients with NHD (1, 32). If these results are confirmed in a much larger sample of patients, sTREM2 concentrations in CSF potentially could be used to screen for TREM2 homozygous missense mutations in an analogous way to progranulin mutation carrier screening (3335). Furthermore, in a cross-sectional analysis of sTREM2 concentrations in CSF of FTD patients, we found significantly reduced concentrations of sTREM2 in FTD patients compared to neurologically normal controls, further supporting a TREM2 loss-of-function mechanism in the pathogenesis of AD and FTD. Although the overlap between groups precludes the current utilization of sTREM2 in clinical settings, we believe that CSF sTREM2 deserves further research as a potential marker for neurodegenerative diseases, probably in combination with other yet to be identified markers. Finally, our findings suggest that TREM2-dependent phagocytosis could be modulated, hence opening the door for new therapeutic strategies. Indeed, as a proof of principle, we showed that this can be achieved by blocking ADAM proteases (Fig. 7). However, therapeutic application would require a search for compounds that selectively and specifically enhance TREM2 activity without interfering with the expression of other proteins or essential signaling pathways.


Study design

The first part of this study was designed to determine whether TREM2 missense mutations, which recently have been identified as risk factor for several neurodegenerative diseases, alter maturation and processing of TREM2 and ultimately cause a loss of TREM2 protein function. We investigated TREM2 processing and function using several cell lines including HEK293 Flp-In cells, microglial cell lines (BV2 cells), and primary microglial cells derived from Trem2−/− mice. The second aim was to study the sTREM2 concentrations in CSF and plasma of patients diagnosed with FTD-like syndrome, FTD, or AD and compare them to those of healthy controls. For these purposes, we performed a cross-sectional study in which we measured sTREM2 in CSF and plasma samples from six specialized neurological referral centers. We studied all samples available from these centers; we did not perform a priori calculation of sample size. Clinical diagnoses were made according to internationally accepted criteria. In the final analysis, we excluded those control individuals and FTD patients who had an AD CSF profile defined by the Mattsson et al. equation (36), and, conversely, we excluded those clinically diagnosed AD patients who did not have an AD CSF profile. sTREM2 measurements were performed in an ELISA-based assay that we developed. The measurements were made in a blinded fashion. To study the consequences of TREM2 loss of function, we performed phagocytosis assays using three independent targets (latex beads, E. coli bacteria, and Aβ1–42) and finally studied whether modulation of TREM2 could alter microglial function.

Cell culture and generation of isogenic cell lines

Flp-In 293 cells (HEK293 Flp-In; Life Technologies) were cultured in Dulbecco’s modified Eagle’s medium (DMEM) with Glutamax I, supplemented with 10% (v/v) fetal calf serum (FCS), Zeocin (200 μg/ml), and penicillin/streptomycin (PAA Laboratories). Transfections of complementary DNA (cDNA) constructs were carried out using Lipofectamine LTX with Plus Reagent according to the manufacturer’s recommendations. For stable overexpression of human TREM2 cDNA constructs, HEK293 Flp-In cells were cotransfected with the TREM2 cDNA constructs and pOG44 (Flp-recombinase expression vector; Life Technologies) and selected using hygromycin B (200 μg/ml). BV2 microglia cells (21) were cultured in DMEM with Glutamax I, supplemented with 10% (v/v) FCS and penicillin/streptomycin, and stable TREM2-expressing cell lines were selected using Zeocin (200 μg/ml). Cell lines were regularly monitored for Mycoplasma contamination via a PCR-based method, and all results were negative throughout the course of the study. If not stated otherwise, products for cell culture experiments were obtained from Life Technologies.

Primary microglia cell culture

Primary microglia were isolated from postnatal day P5 to P6 mouse brains using MACS Technology (Miltenyi Biotec) according to the manufacturer’s instructions. Briefly, brain cortices were dissected, freed from meninges, and dissociated by enzymatic digestion using a Neural Tissue Dissociation Kit P. CD11b-positive microglia were magnetically labeled using CD11b MicroBeads, loaded onto a MACS column, and subjected to magnetic separation. Isolated microglia were plated onto 24-well plate at a density of 8.5 × 104 cells per well and cultured in DMEM/F12 medium (Life Technologies) supplemented with 10% heat-inactivated FCS (Sigma) and 1% penicillin/streptomycin and maintained in a humidified 5% CO2 incubator at 37°C. After 24 hours, cultured medium was replaced with fresh medium for 2 to 4 days until use for phagocytosis assays.

Phagocytosis assays

Bead uptake assays using HEK293 Flp-In cells stably expressing either wild-type or mutant TREM2-DAP12 fusion constructs (25) were performed essentially as described before (23). Briefly, cells were plated in 24-well plates at a density of 5 × 104 cells and cultured overnight. Preopsonized [50% FCS in phosphate-buffered saline (PBS)] latex beads (6 μm, internally dyed with the fluorophore Flash Red; Polysciences Inc.) were added to the cells at a concentration of 20 beads per cell and incubated for 90 min at 37°C. As a negative control, phagocytosis was inhibited with 10 μM cytochalasin D 30 min before addition of latex beads. Cells were harvested using TrypLE buffer (Life Technologies), washed three times with fluorescence-activated cell sorting (FACS) sample buffer (1% FCS and 0.02% sodium azide in PBS), and analyzed by flow cytometry on a FACSCalibur flow cytometer (BD Biosciences). Data analysis was performed using FlowJo version 9.7.1.

Phagocytosis of fluorogenic E. coli particles (pHrodo Green, Molecular Probes) was analyzed using BV2 microglial cells, primary microglia or HEK293 Flp-In cells stably expressing either wild-type or mutant TREM2-DAP12 fusion constructs (25). Briefly, cells were plated in 24-well plates at a density of 2 × 105 (HEK293 Flp-In), 1 × 105 (BV2), or 8.5 × 104 (primary microglia) cells and cultured for 24 to 48 hours. pHrodo E. coli bioparticles were dissolved in PBS to a concentration of 1 μg/μl, and a total of 50 μg of bioparticles was added per condition and incubated for 60 min at 37°C. As a negative control, phagocytosis was inhibited with 10 μM cytochalasin D, which was added 30 min before addition of pHrodo E. coli bioparticles. Cells were harvested by trypsinization, washed two times with FACS sample buffer, and analyzed by flow cytometry on a MACSQuant VYB flow cytometer (Miltenyi Biotec). Data analysis was performed using the MACSQuantify software (Miltenyi Biotec).

Phagocytosis of aggregated FAM-labeled Aβ1–42 (Anaspec) was analyzed similar to previously described methods (37). Briefly, FAM-labeled Aβ1–42 was aggregated overnight at 37°C with agitation. Primary microglia, BV2, or HEK293 Flp-In cells were plated at 2 × 104, 1.5 × 104, 5 × 104 cells per well, respectively, in poly-l-lysine–coated black-walled 96-well plates (Greiner Bio One) and cultured overnight. Aβ was added to a final concentration of 0.5 μM (primary microglia) or 1.5 μM (HEK293 Flp-In and BV2) and incubated for 4 hours at 37°C. Extracellular Aβ1–42 was quenched with 100 μl of 0.2% trypan blue in PBS (pH 4.4) for 1 min. After aspiration, fluorescence was measured at 485-nm excitation/538-nm emission using a Fluoroskan Ascent FL plate reader (Labsystems).

Patient material

CSF samples were obtained by lumbar puncture following standard procedures, collected in polypropylene tubes, and immediately frozen at −80°C until use (36, 38). Six specialized neurological referral centers were involved in the study. AD was diagnosed according to the National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer’s Disease and Related Disorders Association criteria (39). For patients recruited before 2011, FTD was diagnosed according to the Neary consensus criteria (40). Thereafter, patients diagnosed of bvFTD (behavioral variant FTD) followed the new revised bvFTD criteria (41), and those with semantic dementia or progressive nonfluent aphasia fulfilled the primary progressive aphasia international consensus criteria (42). The control group (n = 88) consisted of individuals fulfilling the following inclusion criteria: (i) no neurological or psychiatric antecedents, (ii) no organic disease involving the central nervous system after extensive clinical examination, and (iii) cognitive deterioration was ruled out after evaluation by a neurologist with experience in neurodegenerative dementias. Among them, there were two patients who were diagnosed with disorders of the peripheral nervous system and two with depression. To improve the accuracy of the diagnosis, we excluded those control individuals (6 of 94, 6.4%) and clinically diagnosed FTD patients (7 of 57, 12.3%) who had an AD CSF profile defined by the Mattsson et al. equation (36) [(Aβ42/P-tau)/(3.694 + 0.0105 × T-tau)], and, conversely, we excluded those clinically diagnosed AD patients who did not have an AD CSF profile following the former equation (24 of 80, 30%). The decrease in sTREM2 concentrations in AD and FTD patients compared to control individuals was observed both including and excluding those patients. Among the FTD group, two patients carried a C9orf72 repeat expansion and three subjects carried a progranulin (GRN) mutation. Again, the decrease in sTREM2 concentrations in FTD patients compared to control individuals was observed both including and excluding the genetic cases from the FTD group. All patients gave written informed consent, and the study was approved by the local ethics committees of the participating centers.


To quantify the concentrations of sTREM2 in human CSF, EDTA-plasma samples, or cell culture supernatants, an ELISA for human sTREM2 was established using the Meso Scale Discovery SECTOR Imager 2400 similarly to previously established ELISAs (43). Streptavidin-coated 96-well plates were blocked overnight at 4°C in blocking buffer [0.5% bovine serum albumin (BSA) and 0.05% Tween 20 in PBS (pH 7.4)]. For the detection of human sTREM2, plates were incubated for 1 hour at room temperature with biotinylated polyclonal goat anti-human TREM2 capture antibody (0.25 μg/ml) (R&D Systems) diluted in blocking buffer. Plates were washed subsequently for four times with washing buffer (0.05% Tween 20 in PBS) and incubated for 2 hours at room temperature with samples diluted 1:4 in assay buffer [0.25% BSA and 0.05% Tween 20 in PBS (pH 7.4)] supplemented with protease inhibitors (Sigma). A recombinant human TREM2 protein (Hölzel Diagnostika) was diluted in assay buffer in a twofold serial dilution and used for the standard curve (concentration range, 4000 to 62.5 pg/ml). Plates were washed three times for 5 min with washing buffer before incubation for 1 hour at room temperature with mouse monoclonal anti-TREM2 antibody (1 μg/ml) (Santa Cruz Biotechnology; B-3) diluted in blocking buffer. After three additional washing steps, plates were incubated with a SULFO-TAG–labeled anti-mouse secondary antibody (1:1000; Meso Scale Discovery) and incubated for 1 hour in the dark. Last, plates were washed three times with wash buffer followed by two washing steps in PBS and developed by adding Meso Scale Discovery Read buffer. The light emission at 620 nm after electrochemical stimulation was measured using the Meso Scale Discovery SECTOR Imager 2400 reader.

Spike recovery, linearity, interplate, and interday variability for the human sTREM2 ELISA was determined using both a dedicated CSF and plasma sample (tables S2 and S3). Repeated freeze-thaw cycles had only minimal effects on sTREM2 concentrations in CSF (fig. S4C) and no effect on sTREM2 concentrations in plasma (fig. S4D). The specificity of the used ELISA system was further validated by anti-TREM2 immunoblotting showing high degree of correlation between the ELISA readings and immunoreactivity on the immunoblot using an independent anti-TREM2 antibody (fig. S4F).

To measure murine sTREM2, the same procedure as outlined above was followed using biotinylated polyclonal sheep anti-mouse TREM2 (0.25 μg/ml) as capture antibody, rat monoclonal anti-mouse TREM2 (1 μg/ml) as detection antibody (both R&D Systems), and a SULFO-TAG–labeled goat anti-rat secondary antibody (1:1000; Meso Scale Discovery) as secondary antibody. As standard, a recombinant mouse TREM2 protein (Hölzel Diagnostika) was used.

To quantify the levels of sTREM2 secreted from HEK293 Flp-In or BV2 cells, conditioned media from biological replicates, collected as described above, were analyzed in duplicates using either a commercial TREM2 ELISA according to the manufacturer’s recommendations (Sino Biological; Figs. 1B and 2B) or our newly established human (Fig. 4, A and B) or mouse (Fig. 7, A and B) sTREM2 ELISA. The sTREM2 standard curves were generated using the MasterPlex ReaderFit software (MiraiBio Group, Hitachi Solutions America) through a five-parameter logistic fit.

Statistical analysis

The χ2 test was used to compare differences in categorical variables. One-way ANOVA followed by Tukey post hoc test was used to compare normally distributed continuous variables. Data that did not follow a normal distribution (including sTREM2 concentrations in CSF and plasma) were analyzed with nonparametric tests (Kruskal-Wallis followed by post hoc Mann-Whitney U test). To control for the effect of potential confounders (gender, age, and center origin of the sample) on sTREM2 in CSF, we log-transformed this variable to achieve a normal distribution, and performed an ANCOVA. Statistical significance was set to 5% (α = 0.05). All tests were two-sided, and all data were analyzed using the Statistical Package for the Social Sciences 20.0 (SPSS Inc.).


Materials and Methods

Fig. S1. Absence of TREM2 fragments containing C-terminal FLAG tag.

Fig. S2. Analysis of maturation of TREM2.

Fig. S3. Characterization of BV2 cells stably overexpressing human TREM2.

Fig. S4. Characterization of novel sTREM2 ELISA.

Fig. S5. Impaired phagocytosis by mutant TREM2.

Table S1. Primers used for RT-PCR analysis.

Table S2. Characterization of novel sTREM2 ELISA.

Table S3. Spike recovery and linearity test for CSF and plasma sTREM2 ELISA.

References (4448)


  1. Acknowledgments: We thank R. Guerreiro and J. Toombs. We thank B. Schmidt (Technical University of Darmstadt) for providing the inhibitor GI 254023X and P. Saftig (University Kiel) and Galderma for providing the inhibitor GL 506-3. We thank J. McCarter for critically reading the manuscript. We thank the personnel of the VIB Genetic Service Facility for the genetic screenings and the Antwerp Biobank at the Institute Born-Bunge for the autopsied brain and CSF samples. We also acknowledge the contribution of the clinical neurologists and the research nurses to the biosampling of the patients and the control individuals. Funding: This work was supported by the European Research Council under the European Union’s Seventh Framework Program (FP7/2007-2013)/ERC Grant Agreement No. 321366-Amyloid (advanced grant to C.H.), the Deutsche Forschungsgemeinschaft (German Research Foundation) within the framework of the Munich Cluster for Systems Neurology (EXC 1010 SyNergy), the general legacy of Mrs. Ammer (to the Ludwig-Maximilians University/the chair of C.H.), and the Leonard Wolfson Institute for Experimental Neurology (to J.H.). This work was partly funded by the Instituto de Salud Carlos III (FISPI11/3035) to A.L. The Antwerp site was in part funded by the Belgian Science Policy Office Interuniversity Attraction Poles Program, the Flemish government–initiated Methusalem Excellence Program, the Alzheimer Research Foundation, the Medical Foundation Queen Elisabeth, the Research Foundation Flanders (FWO), the Agency for Innovation by Science and Technology Flanders (IWT), and the University of Antwerp Research Fund, Belgium. The IWT provides a PhD fellowship to E. Cuyvers, and the FWO provides a postdoctoral fellowship to J.v.d.Z. Author contributions: C.H. set up the research concept. G.K., M.S.-C., and C.H. designed the experiments and wrote the manuscript. G.K. and N.P. carried out all experiments except the TREM2 ELISA (fig. S3E), which was performed by Y.Y. under the supervision of M.C., and the phagocytosis assay with latex beads, which was carried out by E. Czirr under the supervision of T.W.-C. F.M. characterized cell lines and A.W.-W. and S.T. isolated and cultured primary microglia. J.H., M.C., M.S.-C., M.W., and S.L. provided valuable conceptual advice. M.S.-C. performed statistical analyses. E. Cuyvers, K.S., J.v.d.Z., and C.V.B. recruited patients, relatives, and control individuals; performed the genetic screening of AD and FTD genes; and provided mutation and clinical data of TREM2 mutation carriers. M.S.-C., E.L., H.S., A.L., D.A., J.F., J.L.M., J.-J.M., R.S.-V., A.A., A.R., M.T.H., A.D.-T., S.E., H.Z., and H.G. recruited patients and control persons and provided CSF samples. J.-J.M. performed autopsy diagnosis and provided neuropathology and immunohistochemistry data of AD and FTD patients. Competing interests: The authors declare that they have no competing interests.
View Abstract

Navigate This Article