Erythromyeloid-Derived TREM2: A Major Determinant of Alzheimer’s Disease Pathology in Down Syndrome

Subjects

Human brain tissues from controls (age 60±15 years), DS (age 60±15 years), and AD (age 82.0±8.0 years) (n = 6 in each group) were provided by the Cambridge Brain Bank (Table 1). For use of human brain tissue and serum samples, The Cambridge Health Authorities Joint Ethics Committee granted ethical approval. Whole blood and serum samples from human controls (n = 50), DS (n = 50), and AD (n = 50) were collected for DNA and protein analysis. Written consents were obtained from typically developing controls and all adults with DS with capacity to consent. Verbal assent was obtained from participants with DS lacking capacity and a written assent was provided by an appointed consultee, in accordance with the UK Mental Capacity Act (2005). Ethics and research and development approvals were granted by the National Research Ethics Committee of East of England –Norfolk and Cambridgeshire and Peterborough NHS Foundation Trust, respectively.

Blood collection and blood and bone marrow slide preparation

We analyzed 50 cases of DS diagnosed at our center during a three-year period from July 2012 to June 2015. Blood samples were collected into serum blood collection tubes and serum immediately separated by centrifugation at 2,465 g for 6 min at 4°C, aliquoted, and stored at –80°C until analysis.

A drop of blood or bone marrow placed on a slide was spread using a spreader as described in Dacie and Lewis [26]. The slide was air-dried and fixed in 100% methanol and stored at 4°C. Further investigations, including immunohistochemistry (IHC), were performed on stored fixed slides as described later.

Biochemical and hematological profiles were analyzed by pathology laboratories at Addenbrooke’s Hospital, Cambridge University Hospitals NHS Foundation Trust, Cambridge.

Genotyping

Single nucleotide polymorphism (SNP) genotyping for TREM2 SNP, rs75932628, C-T encoding the R47H variant was performed using the TaqMan^® allelic discrimination assay on a HT7900 sequence detection system (Applied Biosystems), according to the manufacturer’s instructions. Genotyping success rates were >96%. There were no inconsistencies amongst 92 samples genotyped in duplicate. APOE (ɛ2,/ɛ3 and ɛ4) genotype was analyzed by Genotype Facilities, East Anglia Medical Genetics Service, Addenbrooke’s Hospital, Cambridge UniversityHospitals NHS Foundation Trust, Cambridge.

Cell culture

Antibodies

The following primary antibodies were used mouse monoclonal (MAB) anti-TREM2 (ab201621, MM0942-42E14) and rabbit monoclonal (ab209814) and other antibodies (Supplementary Table 1) from Abcam (Cambridge, UK). The monoclonal anti-Aβ₄₂ antibodies (6E10) (Signet laboratory) have been described previously [21, 28] and other antibodies used in this study are shown in Supplementary Table 1. The following secondary antibodies were used: biotinylated goat anti-rabbit and biotinylated horse anti-mouse (both from Vector Laboratories, 1 : 250 for IHC); Alexa Fluor 568-labeled donkey anti-mouse, Alexa Fluor 488-labeled donkey anti-rabbit, and Alexa Fluor 568-labeled donkey anti-goat (all from Invitrogen, 1 : 1000 for immunofluorescence).

SDS-PAGE and western blotting

The HEK293T cell line was transfected with human TREM2 cDNA (Human cDNA ORF clone, NM_018965). Cell lysates from wild type HEK293 cells and TREM2 transfected HEK293 cells were prepared. Similarly, protein lysates were prepared from the temporal cortex and hippocampus of human brains (n = 6), and human serum (n = 50 in each group). 20μg protein samples were separated on 4–12% Nu-PAGE^® Bis-Tris (Bis (2-hydroxyethyl)-amino-tris (hydroxymethyl)-methane) gradient gels (1.25 M Bis-Tris pH 6.4, 30% acrylamide/bis), run with NuPAGE^® MES SDS running buffer (Lifetechnologies), and transferred to 0.45μm pore size polyvinylidene difluoride (PVDF) membranes (Invitrogen). Membranes were incubated with the appropriate primary antibody in blocking buffer (1% milk in 0.1 M tris saline buffer containing 0.1% Tween 20) for 24 h at 4°C and then washed three times with 0.1 M tris saline buffer containing 1% Tween 20 (TBST) followed by incubation for 1 h at room temperature with HRP-conjugated secondary antibodies (anti mouse IgG 1 : 3000, DAKO) or anti-rabbit IgG (1 : 3000; DAKO) antibodies). Binding was detected with ECL Plus chemiluminescence reagents and Hyperfilm ECL (both from GE Healthcare). Immunoblot controls consisted of pre-adsorption of the antibody with an excess of the immunogenic peptides (Abcam Ab174422), that blocks the anti TREM2 antibody used in the study (Abcam 201621), resulting in the absence of immunoreactive bands in the membrane.

Immunofluorescence

Brain sections, blood smears, or cells were blocked using blocking buffer (0.1 M PBS, 0.3% Triton X100, 10% normal donkey serum) for 1 h at room temperature, then incubated overnight at 4°C with primary antibody diluted in blocking buffer. Alexa Fluor-conjugated secondary antibodies were used for detection and samples counterstained with4^′6-diamidino-2-phenylindole (DAPI, Sigma). Sections were then mounted on glass slides with coverslips using FluorSave (Calbiochem).

Microscopy

Bright field images were taken and quantified using Leica imaging software and a Leica FW 4000 upright microscope equipped with a SPOT digital camera. Fluorescence images were obtained using a Leica DM6000 wide field fluorescence microscope equipped with a Leica FX350 camera with x20 and x40 objectives. Images were taken through several z-sections and de-convolved using Leica software. A Leica TCS SP2 confocal laser-scanning microscope was used with x40 and x63 objectives to acquirehigh-resolution images.

Image and statistics analysis

All sections for (IHC and WB) were performed in triplicate. WB images were quantified using ImageJ software (US National Institute of Health), normalizing all samples to loading controls. Values in the figures are expressed as mean±SEM. To determine statistical significance, values were analyzed by Student’s t-test. A probability value of p < 0.05 was considered to be statistically significant.

Soluble TREM2 protein level declined in DS serum with disease progression

As a first step towards determining possible roles of TREM2 in DS, we investigated TREM2 protein levels in human brain tissue and serum by WB using anti-TREM2 mouse monoclonal antibody (MAB) (ab201621, Supplementary Table 1).

To validate that TREM2 protein was recognized by TREM2 antibody (ab201621), cell lysates obtained from the HEK293 cell line transfected with Human TREM2 cDNA was analyzed by WB. The antibody recognized bands of 25 kDa and ∼18 kDa, respectively, in the cell lysates of these TREM2 overexpressing HEK293 cells, and in control HEK293 cell lysates no bands were seen, as expected (Fig. 1a). To investigate TREM2 protein expression in human brain, tissue from the superior frontal cortex of AD, DS, and age-matched controls brains (n = 6 in each group), TREM2 protein was detected as non-glycosylated 25 kDa and ∼18 kDa truncated bands. There was an additional ∼50 kDa band visible which may potentially be a TREM2 dimer (Fig. 1b). The TREM2 levels (i.e., the 25 kDa bands) were highest in control, lower in DS and lowest in AD brains (Fig. 1b, 1c; p < 0.001).

Fig.1

TREM2 protein levels declined with Alzheimer’s diseases progression in Down syndrome. To validate the TREM2 antibody (ab201621), the HEK293 cell line was transfected to overexpress human TREM2, cell lysates were analyzed by WB. The antibody recognized a band of 25 kDa and ∼18 kDa in the TREM2 overexpressing HEK293 cell lysate, whereas in control HEK293 cell lysate no band was seen (a). The WB analyses performed on postmortem human brain tissue (superior frontal cortex) showed that TREM2 levels were highest in controls, significantly decreased in DS and lowest in AD brains (b & c, p < 0.0005). TREM2 serum levels were significantly increased in young DS, lower in middle age, and lowest in the older subjects (e & f, p < 0.001 between older versus younger DS). Similarly, serum TREM2 levels were lower in AD (n = 25), compared with age-matched controls (n = 25) and lowest in DS (h & i, p < 0.001). β-actin and human serum albumin were used as positive controls for brain tissue and serum samples, respectively (d, g, & j). Bonferroni-corrected Student’s t-tests shown; (n = 8–25 per group). Error bars indicate SEM. ^*p < 0.05, ^**p < 0.01, ^***p < 0.001, one-way ANOVA.

Serum samples obtained from DS, AD, and controls (living subjects) were analyzed by WB. We divided DS subjects into three age groups (younger DS, n = 16, 30–39 years; middle aged DS, n = 20, 40–49 years; and older DS subjects, n = 8, 50–65 years— none of the younger DS subjects up to age 45 had developed clinical dementia). WB showed that samples from younger DS subjects had ∼50 kD and 25 kDa bands and a multiple banding pattern between ∼7–18 kDa weights, which were possibly due to truncation of the protein as a result of ectodomain shedding, as described previously [29] (Fig. 1e). More truncated TREM2 bands were detected in DS serum compared to controls. Serum TREM2 levels were significantly higher in younger DS, lower in middle aged DS and lowest in the older subjects (Fig. 1e, f; p < 0.001 when comparing older versus younger DS). Similarly, serum TREM2 levels were lower in AD (n = 25, p < 0.001) when compared with age-matched controls (n = 25) but lowest in DS (Fig. 1h, i; p < 0.001). Anti-β-actin for brain tissues and human anti-serum albumin antibody for serum were used as protein loading controls(Fig. 1d, g, j).

These data indicate that further truncation of serum TREM2 following ectodomain shedding may be age or disease related. Furthermore, in DS and AD brain tissue, our results show that TREM2 protein levels decrease in AD brain with age and diseaseprogression.

TREM2 protein is absent in senile plaques in DS brain

We next attempted to visualize TREM2 expression using immunofluorescent staining on sections of postmortem brains superior frontal cortex and temporal cortex from AD, DS, and age-matched controls (n = 6 in each group: 46 to 76 years, Table 1). Extensive variation of TREM2 expression was observed in different brain regions as well as between DS cases. In the superior frontal cortex of control brain, anti-TREM2 and anti-amyloid-β (Aβ₄₀) co-localized in pyramidal neurons and in small vesicles (Fig. 2a-c). In DS brain sections, although many mature senile plaques (SPs) with well-defined cores of amyloid-β-42 (using Aβ₄₂, 6E10 antibody) protein were observed, only small numbers of cells in thecenter of the SPs were positive for TREM2 (Fig. 2d-g). In one DS case (DS6, 66 years old), only a very limited amount of TREM2 protein was present in the hippocampus (molecular grey matter) adjacent to the dentate gyrus, and there was increased signal intensity around the cortical neuron (Fig. 2f). Whereas in another sample from a subject (DS2, 76 years old) showed Aβ₄₂ positive plaques but there was no TREM2 expression visible (Fig. 2g). Similarly, in AD brain sections stained with TREM2 and Aβ₄₂, TREM2 protein was not detected in SPs and only found peripheral to the plaques (Fig. 2h-j). TREM2 protein expression was higher in control brains, lower in AD and lowest in DS brains (Fig. 2k; p≤0.001). High power DS brain images were captured with confocal microscopy. In a young DS subject (DS3, 46 years old), TREM2 protein was visible in the layer V pyramidal neurons of the frontal cortex (Fig. 3a), and in the periphery of SP (Fig. 3b-d). We have previously reported that TREM2 protein is expressed in dentate gyrus granule cells of wild type mice [21]. TREM2 protein was also observed in the cerebellar granule cell layer in DS brain (Fig. 3e, f). These data suggest that TREM2 occurs in young healthy pyramidal neurons and in granule cells, but there is only very limited expression in damaged (diseased) neurons and it does not co-localize withAβ₄₂-positive SPs.

Fig.2

TREM2 protein is absent in senile plaques in DS brain. Double immunofluorescence staining was performed in the temporal cortex of normal control brain tissue by using the rabbit polyclonal anti-Aβ₄₀ (green), mouse monoclonal anti-TREM2 (red), and DAPI for nuclei (Blue). Aβ₄₀ immunoreactivity was visible in pyramidal neurons (a), TREM2 in the cell body (b), and both proteins co-localized in the perinuclear location in pyramidal neurons (c). In DS (DS6), layer III temporal cortex sections stained for Aβ₄₂ (green) and TREM2 (red) showed many mature plaques (d & e). TREM2 was rarely present in the core of the plaques (d) and in neurons (f). Further staining on a different case (DS2) showed that Aβ₄₂ did not co-localize with TREM2 (g). In AD brain, TREM2 protein was visible around the plaques (h-j). Aβ₄₂ level was highest in AD brain (k, p < 0.0001) and TREM2 level was lower in surviving neurons compared to control and AD brain (K, p < 0.01). The scale bar in a-e = 50μm, f = 25μm, and g = 40μm. These data represent the mean±S.E from three independent experiments. Error bars indicate SEM. ^*p < 0.01, ^**p < 0.001, ^***p < 0.0001.

Fig.3

Confocal images of senile plaques and cerebellar granule cells in DS brain. Confocal microscopy analysis was performed on DS brains sections, using anti-Aβ₄₂, anti-TREM2 and counterstained with DAPI (blue). DS brain sections from frontal cortex revealed TREM2 protein was visible in the layer V pyramidal neurons (a), characteristic neuritic plaques stained with Aβ₄₂ (b-d), and limited TREM2 expression was seen in the SP (c, d). A high-power image showed TREM2 (green) appeared to be expressed in the cerebellar granule cells (e, f). Scale bar in a-d = 25μm, e = 50μm, f = 15μm.

TREM2 expression in a selective population of phagocytotic microglia and macrophages in DS and AD brains

Recent work has shown TREM2 to be involved in microglial activity [30, 31] and the amyloid clearance process. We therefore investigated TREM2 expression in microglia, macrophages and astrocytes in control, AD and DS brains. In control brains, we found no TREM2 co-localization with the astrocytic marker GFAP (Fig. 4a). However, in AD brain sections GFAP positive astrocytes were present around SPs and they appeared very granular and damaged (Fig. 4b). In DS brain sections, activated astrocytes were visible close to the SPs but no co-localization with TREM2 was observed (Fig. 4c, individual panel was shown in Supplementary Fig. 1a-c). We measured the relative intensity of TREM2 and GFAP and although there was no difference in total TREM2 levels around astrocytes, there were higher numbers of activated astrocytes in AD brains compared to DS and controls (Fig. 4d; p≤0.001).

Fig.4

TREM2 expression in a selective population of phagocytotic microglia and macrophages in DS and AD brains. Brain sections were stained with the astrocytes marker GFAP (green, indicated with arrow in astrocytes) and TREM2 (red, with arrow head) (a-c). In control brain section, TREM2 was seen in the in the neurons surrounded by GFAP positive astrocytes (a), in AD brain sections astrocytes were granular, fragile and seen close to the plaques (b), whereas in DS brain activated astrocytes were present in the vicinity of the amyloid plaques (c). Highest expression of GFAP positive astrocytes was found in AD brain (d, p < 0.001). Iba1 positive ramified microglia were seen in control brain (e-g), while activated Iba1 (green) positive microglia were visible in DS brain (h) and a TREM2 positive macrophage was visible on pial surface and small vesicular particle (i), co-localized with TREM2 and Iba, showed with arrow head (j). In AD brains, a small number of Iba1 positive microglia co-localized with TREM2 close to senile plaques (k, l), and in hippocampus (m), indicated with arrow in microglia and arrow head in neurons. There were more microglia positive for Iba1 in control brain compared to DS and AD (n, p < 0.0001). In the pial surface and in blood vessels, TREM2 (green) was visible in peripheral macrophages where it co-localized with the macrophage marker CD68 (red) (o-q). In DS brain, strings of blood cells were visible in the brain parenchyma that co-localized with TREM2 (q) and highly significantly (r, p < 0.0001). Scale bar in a-c = 25μm, e-g & q-U = 50μm, h-j = 20μm. Image intensity measured by Image J (d, n, and r). Error bars indicate SEM. ^*p < 0.01, ^**p < 0.001, ^***p < 0.0001.

We then stained brain sections with microglial marker (Iba1) and TREM2. In control brains, Iba1 positive microglia were visible in the cortex(Fig. 4e-g; Supplementary Fig. 2a-c). TREM2 was present in some cells close to blood vessels but there was limited TREM2 expression in the ramified microglia (Fig. 4f, g; Supplementary Fig. 2a-c). In DS brains, TREM2 protein was found in surviving cells around the SP (Fig. 4h; Supplementary Fig. 1d-f). Iba1 positive activated microglia were present in the brain parenchyma but there was limited co-localization with TREM2 (Fig. 4h; Supplementary Figure 1d, f). In DS brain, soluble TREM2 was observed to be present in a cell on the pial surface of a macrophage phenotype in addition to a subset of microglia within the cortex (Fig. 4i, j; note also small vesicles shown by arrow head). In AD brain, activated microglia were present near SPs in the cortex (Fig. 4k-m; Supplementary Figure 2d-f) and limited numbers of damaged microglia co-localized with TREM2 in the periphery of SP (Supplementary Figure 2f) and in the hippocampus (Fig. 4m; p≤0.01). The highest number of TREM2 positive cells was seen in control brains and the lowest in DS brains (Fig. 4n; p≤0.001).

As TREM2 protein was visible in peripheral macrophages in post mortem brain tissue, we therefore extended our search using anti-CD68 (phagocytotic macrophage/microglia marker), and anti-TREM2 antibodies. In all brain sections, TREM2 showed very strong co-localization with CD68 (Fig. 4o-q, r; p≤0.001). In DS brain, red blood cells (RBCs) were visible on the pial surface of the cortex and TREM2 protein was found inside the brain parenchyma (Fig. 4q). These data support the notion that macrophages and red blood cells from blood vessels can transport soluble TREM2 protein into the brain parenchyma.

TREM2 protein observed in the brain parenchyma close to blood vessels

To confirm the location of TREM2 in cells related to peripheral blood vessels in the brain parenchyma, we investigated the expression of DAP12 and the macrophage marker CD68. DAP12 is a type-I transmembrane adapter protein that forms a molecular complex with various receptors, including TREM2 [32]. In control brain sections, TREM2 and DAP12 were visible in neurons in the brain parenchyma (Supplementary Figure 3a-c). To confirm neuronal expression, brain sections were stained with TREM2 and a neuronal marker (NeuN), which were found to co-localize (Supplementary Figure 3d). In DS and control brain sections, TREM2 was found in perivascular macrophages close to blood vessels (Fig. 5a-c). DS brain sections were stained with anti-TREM2 and anti-CD68, and were found to co-localize (Fig. 5d). TREM2 positive cells were observed to be in direct contact with the pial surface and only a small subset co-localized with Iba1 positive cells (Supplementary Figure 3e-f). DS brain sections close to the blood vessels were stained with anti-Aβ₄₂ and anti-TREM2 and imaged with confocal microscopy. Aβ₄₂ was visible in the brain parenchyma and blood cells whereas TREM2 was seen in cells in close proximity to blood vessel walls (Fig. 5e-g).

Fig.5

TREM2 protein enters in the brain from blood vessels via macrophages. Immunofluorescence staining was performed with polyclonal rabbit anti-DAP12 (green), monoclonal anti-TREM2 (red) antibodies, and DAPI (Blue). In control brain sections, DAP12 was visible in the brain parenchyma (a), whereas TREM2 was found in the perivascular macrophages and close to the blood vessels in DS brain (b). In DS brain a TREM2 positive macrophage was seen in blood vessels (c). DS brain sections were stained with TREM2 and CD68, showed that TREM2 protein carried by RBCs (d). A confocal micrograph showed Aβ₄₂ (green) was visible in the parenchyma and in the blood cells whereas TREM2 (red) was seen in the macrophages close to the blood vessels (e-g). Scale bar in a-b = 70μm, c-d = 50μm, and e-g = 20μm.

Haplotype co-relation between TREM2 and APOE in DS

We performed single nucleotide polymorphisms (SNPs) analysis for the TREM2 (R47H, C/T) mutation and APOE genotype (ɛ2, ɛ3, and ɛ4 alleles) in our DS DNA from blood samples (n = 50). The frequency of the TREM2 R47H (T allele) mutation was 0.046-only two subjects out of 44 carried the mutation; a similar frequency was previously reported in AD and older population [16]. The frequency of APOE alleles ɛ2 (0.11) and ɛ4 (0.15) were greater in DS, whereas ɛ3 (0.74) was lower compared to previously published studies on AD (ɛ2, 0.066; ɛ3, 0.85; and ɛ4, 0.08) [33]. None of DS subjects were homozygous for the APOE4 allele. One DS subject (DS55, 39 years old) with TREM2 R47H (T allele) was heterozygous for APOE (ɛ3/ɛ4) and another DS subject (DS18, 32 years old) was homozygous for APOE (ɛ3/ɛ3). Both of them did not have clinical dementia, and may have been too young to show clinical symptoms. There was a positive co-relation between the TREM2-C allele and APOE ɛ2 allele as a protective haplotype, but the number of DS samples available to study were too limited to infer any statistical significance.

TREM2 was observed in erythromegakaryocytes and in platelets, suggesting a myeloid origin

There is an increased risk of leukemia in adults with DS in addition to a variety of other hematological abnormalities that have been reported [9]. Therefore, we analyzed TREM2 expression in blood smears from DS subjects (n = 50, from different age groups 30–65 years) and imaged with confocal microscopy. DS and control blood smears were co-labeled with TREM2 and an erythrocyte plasma membrane marker glycophorin. Both proteins were found to co-localize in RBCs membranes suggesting that TREM2 was expressed on surface of RBCs (Fig. 6a-f). There was stronger TREM2 expression around the RBC plasma membranes, in platelets and in neutrophil plasma membrane in the control samples (Fig. 6a-c) compared to DS samples (Fig. 6d-f, o; p≤0.001). Additionally, TREM2 was observed scattered and extra-corporally to RBCs and within the cell bodies of mononucleated cells (MNC) in DS samples (Fig. 6d-f). A classic dysmorphology (bi-lobed neutrophils) with significantly higher TREM2 protein around the MNC was frequently observed in young DS blood smears (Fig. 6h-j), whereas normal neutrophils positive for TREM2 were found in controls (Fig. 6 g). The blood smear from a young DS subject (DS1, 30 years old) showed more TREM2 protein expression within RBCs (but not around the erythrocyte plasma membrane) and around the neutrophils (Fig. 6d, h). Two young DS participants’ blood sample smears (DS18, aged 32 years and DS55, aged 39 years, with the TREM2 R47H, T mutation) showed abnormally shaped RBCs in their smears (Fig. 6e, f; and see below), with abnormal accumulation of TREM2 around the MNCs (Fig. 6i, j). DS55 was heterozygous for APOE genotype (E3/E4) and displayed an abnormal erythrocyte phenotype compared to DS18 (ɛ3/ɛ3 genotype) (Fig. 6i). Another blood smear of a young DS participant (DS2, 30 years old negative for R47H mutation and carrying APOE (ɛ3/ɛ4 genotype) when stained with TREM2 and APOE or Aβ₄₂, showed that both proteins were co-localized in platelets and around the MNC (Fig. 6k, m, p, q; p≤0.001). A confocal micrograph of blood smear from DS2 showed large numbers of small particle (like vesicles or exosomes) around the platelets and MNCs. To confirm this, we stained some other slides with exosome markers CD9 or CD63 and TREM2, and both co-localized, indicating that TREM2 could be transport by plasma exosomes particles (Fig. 6n).

Fig.6

TREM2 was observed in erythromegakaryocytes and in the platelets suggesting a myeloid origin. Confocal micrograph images taken of DS and control blood smears, were stained with anti-TREM2 (green) and anti-glycophorin (red), a red blood cell (RBC) membrane marker. TREM2 protein (green) was present around the erythrocytes membrane (a-c). There was a stronger TREM2 (green) expression around the RBCs in the control blood smear (a) than in DS (d-f). Additionally, analyses on the DS blood smears revealed the presence of scattered TREM2 (green) outside mononucleated cells (g-j). In two young DS subjects that were carrying TREM2 C/T variant (R47H mutation, DS18, and DS55), blood smears contained abnormal shape of the (i-j) and soluble TREM2 and APOE around the neutrophil (k). In mutant carrier DS 55 (mutant carrier) showed abnormal shape RBCs and a very faint TREM2 expression was seen in the center of RBCs and membrane bound Aβ₄₂ (green) was visible (l). In young DS2 (APOE4 carrier) carried fragmented platelets surrounded by vesicles positive for TREM2 (m) and co-localized with exosome marker CD9 (n). Bar diagrams: TREM2 level was lower in mutant DS (o), Aβ₄₂ and APOE levels was higher in older DS (p & q, ^***p < 0.0001) Scale bar in a-f, k&l = 20μm, g-j = 10μm, m&n = 5μm.

RBCs appeared macrocytic and both of the DS participants (DS18 and DS55) who were positive for the TREM2 variant (R47H mutation, C/T variation), had significantly higher mean red-cell volumes and low hemoglobin, white cell (3.6×10³/L) and platelet counts (50×10⁹/L), and showed abnormal shape RBCs (Fig. 6l, p≤0.001). These findings indicated that TREM2 protein may be essential for normal development of RBCs and for transporting soluble proteins.

TREM2 is expressed in the human bone marrow myeloid cells and mesenchymal stem cells and it participates in erythro-phagocytosis

To confirm that TREM2 is expressed in human bone marrow megakaryocytes, a DS bone marrow slide was stained with TREM2 and a well-known megakaryocyte marker CD42b. TREM2 was expressed in bone marrow megakaryocytes (Fig. 7a). Human primary mesenchymal stem cells (MSC) were stained with TREM2 and CD42b, which showed that TREM2 was present in human MSC (Fig. 7b), and both proteins co-localized in a human monocyte/myeloid cell line, THP1 (Fig. 7c). As RBCs were engulfed by microglia/macrophages in human brain sections as seen in Fig. 4p and q, we investigated whether TREM2 was involved in erythrocyte phagocytosis using the THP1 cell line (Fig. 7c, d). THP1 cells were treated with freshly harvested IgG coated rat RBCs (Fig. 7e-g), which was previously described method used to investigate RBC phagocytosis [27]. We confirmed that TREM2 protein was present in RBCs, which then participated in phagocytosis and endocytosis. RBCs were engulfed by the THP1 cells within 4 h and membrane bound TREM2 was visible (Fig. 7e-g) and TREM2 co-localized with EEA1 (an early endosome marker) and LAMP1, (a late endosomal marker) (Fig. 7h, i). Cells were harvested and TREM2 protein analyzed by WB showed that the TREM2 levels increased at least four-fold after RBCs treatment suggesting that TREM2 protein was present in RBCs (Fig. 7m, n; p≤0.001).

Fig.7

TREM2 protein is involved in phagocytosis and is required for myeloid cells survival. In human bone marrow, TREM2 (green) protein was expressed in myeloid progenitor cells and co-localized with mesenchymal cell markers anti-CD42b (a). Similarly, to human mesenchymal stem cells (MSC), both TREM2 (red) and CD42b (green) were present (b), and showed some co-localization in monocytic cell line THP1 (c). A bright field image of normal THP1 cells (d) and after 24-h post-treatment with rat RBCs showed total phagocytosis of RBCs inside the THP1 cells (e-f) and co-localization with CD42b (g). RBC-treated THP1 cells were stained with TREM2 and found to be co-localized with EEA1 (green) (an early endosome marker) (h) and LAMP1 (a late endosomal marker) (i), suggesting that TREM2 protein participates in endocytosis and phagocytosis. WB analysis was performed with rat-RBC treated and untreated THP1 cell lysate. TREM2 protein levels increased at least four-fold post-RBC treatment (m & n, p < 0.0001). THP1 cells were transfected with human TREM2 anti-sense oligonucleotides, cells were stained with anti-TREM2 and anti-DAP12 antibodies. Very faint TREM2 expression was visible (j-l). Cells were collected and a WB analysis was performed with an anti-TREM2 antibody. TREM2 expression was visible in the control cells but not in the siRNA treated cells (o & p, p < 0.0001). Error bars indicate SEM. ^***p < 0.0001. The scale bar in a-c and g-i = 25μm, d-f and j-l = 50μm.

Knock down of TREM2 using siRNA in THP1 cells

We have previously reported that TREM2 protein is important for cellular proliferation [21]. To further investigate the role of TREM2 in cellular proliferation, THP1 cells were transfected with human TREM2 anti-sense oligonucleotides for 24–48 h, which resulted in an increase in cell death. After 48 h, cells were stained with TREM2, CD42b, and DAP12 antibodies and only very faint TREM2 expression was visible (Fig. 7j-l). Cells were harvested and a WB analysis performed using mouse anti-TREM2 MAB (ab201621) which showed TREM2 expression in control cells (treated with scrambled SiRNA) but not in TREM2-siRNA treated cells (Fig. 7o, p; p≤0.001).

Conclusion

In this paper, we have confirmed that TREM2 protein originates in megakaryocytes, it is a serum protein transported in blood vessels by erythrocytes, and it migrates across the wall of ventricles to enter brain parenchyma. Individuals with DS show myeloid cell abnormalities from birth that are likely to affect their bone density (osteoclast dysfunction) and myelination, and later in life the lack of functioning TREM2 protein may accelerate neuronal cell loss, contributing to the aging process and neuroinflammation. As first proposed by Hardy and Selkoe, the “amyloid cascade” hypothesis purports that mis-metabolism and subsequent deposition of Aβ peptides as amyloid-β plaques are principle etiopathological events in AD [67, 68]. We propose the “myeloid hypothesis”, an additional conflating mechanism unique to DS where myeloid cell defects due to aneuploidy (impairment of protein transport, myelination and innate immunity) accelerate disease-related processes that lead to earlier onset of the pathology of AD. We have shown that TREM2 plays a critical role in this aging process. This hypothesis would explain why AD pathology occurs some 15 to 20 years earlier in the DS population. Understanding the molecular mechanisms and identifying critically important of TREM2 proteins involved in neuroprotection may reveal novel therapeutictargets.