What this is
- This research investigates the role of the gene variant in () pathology.
- It compares postmortem brain samples from individuals with and without the variant, focusing on amyloid beta (Aβ) and tau pathology, microglial response, and gene expression.
- Findings suggest that variant carriers show no significant pathological hallmarks of despite altered microglial and expression profiles.
Essence
- variant carriers show no Aβ or tau deposition, indicating that other factors may initiate . The study reveals altered microglial and expression profiles in these individuals compared to those with Alzheimer's.
Key takeaways
- No significant differences in Aβ or tau loads were found between variant carriers and sporadic cases, suggesting that does not directly influence these pathologies.
- variant carriers exhibited altered microglial morphology, with increased amoeboid microglia compared to sporadic cases, indicating a potential change in inflammatory response.
- Gene expression analysis revealed 124 genes significantly upregulated in sporadic cases compared to controls, highlighting the complex role of in neuroinflammation and pathology.
Caveats
- The study's sample size is limited, which may affect the generalizability of the findings. Further research is needed to confirm the role of variants in .
- Differences in gene expression and protein profiles were observed, but the functional implications of these changes remain unclear and warrant further investigation.
Definitions
- TREM2: A gene that encodes a receptor involved in immune response, particularly in microglial activation and function in the brain.
- Alzheimer's Disease (AD): A progressive neurodegenerative disorder characterized by memory loss, cognitive decline, and the accumulation of amyloid plaques and tau tangles in the brain.
AI simplified
Introduction
Triggering receptor expressed on myeloid cells 2 (TREM2) was identified as a genetic risk factor for late onset Alzheimer’s Disease (AD) with a similar odds ratio to APOE ε4 in 2012, although varying odds ratios have been found in different populations (1, 16, 18, 25, 28, 33, 43, 49, 52). The frequency of the R47H variant is 0.23‐0.25 depending on different reports with an estimated odds ratio of 4.46 for AD which increases to 4.62 in cases with a European descent (2). TREM2 is a 40kD, 230 amino acid transmembrane protein belonging to the immunoglobulin family that is expressed on the plasma membrane of a number of different dendritic cells, including microglia (24, 27, 42, 44). It is highly conserved and is thought to be a hub or highly connected gene for microglia in a number of different brain regions, including regions that are affected in AD (13, 35). TREM2 is thought to function through two different pathways, one suppressing inflammation and one that aids phagocytosis of any debris in or around neurons (14, 16, 45, 56). Specifically inhibiting the release and secretion of microglial cytokines and releasing tumor necrosis factor (TNF) to promote survival (21). Multiple studies have shown that when TREM2 is downregulated there is less microglial activation, impaired phagocytosis of injured neurons/Aβ plaques and an increase in levels of TNF and nitric oxygen species whereas when TREM2 is overexpressed there is increased phagocytosis and a decrease in pro‐inflammatory signals (24, 54).
The role of TREM2 in AD has been investigated in animal models with amyloid or tau pathology and was found to be upregulated or positively correlated with the presence of amyloid plaques and phosphorylated tau (14, 16, 21, 32, 34, 35, 38, 57). However, Lue et al (32) showed no correlation between TREM2 levels and amyloid plaque burden but did confirm the correlation with tau pathology in human postmortem temporal cortices from AD cases. However, when TREM2 is upregulated the number of microglia surrounding plaques increases or the activation of inflammatory pathways is triggered via activation of tau kinases, leading to greater levels of phosphorylated tau (20, 21, 38, 61).
Numerous studies have implicated TREM2 in Aβ phagocytosis by increasing activation or proliferation of microglia around the amyloid plaques. It has been proposed that a microglia barrier expressing TREM2 restricts amyloid plaque growth protecting neurons from damage (61). Jay et al, demonstrated that in APPPS1‐21; TREM2−/− models at 4 months of age there was reduced Aβ in the hippocampus but no change in the cortical load compared to APPPS1‐21; TREM2+/+ mice and at 8 months of age there was no difference in amyloid load measured in the hippocampus (20). Moreover, 8.5 months of age 5xFAD; TREM−/− mice showed no effect on the Aβ cortical load, but had an increase in hippocampal Aβ deposition (58, 59). In addition, when TREM2 was overexpressed in primary microglia from the same mice at 7 months and 18 months, there was increased Aβ phagocytosis or no response from the microglia, respectively (21). However, the 18‐month‐old microglia were less able to phagocytose Aβ than the 7‐month‐old microglia in normal conditions. The effect of TREM2 on tau deposition has been investigated in P301S mice, showing increased tau pathology and phosphorylation. Whereas overexpression of TREM2 at 7 months rescued this effect by decreased phosphorylation and reduced neuronal and synaptic loss in the hippocampus (21, 22).
Variance in pathology observed using mouse models has led to speculation about the effect TREM2 is playing on inflammatory processes when the TREM2 R47H variant is present. TREM2 in mice has been shown to act upon different mechanisms to human TREM2 variants. The R47H variant reduces TREM2 mRNA expression and splicing but the normal transcripts were observed in human TREM2 R47H (60). Additionally, TREM2 R47H and R62H variants in humans were shown to have a decreased reactive microglial phenotype compared with the TREM2‐dependent mechanisms seen in mice models using single‐nucleus transcriptomics (63). These studies highlight the differences observed between mice and humans and show the need for further studies using TREM2+ variant human post mortem tissue.
APOE was identified as a genetic risk factor for AD, with different allelic compositions determining the level of risk (50). As both APOE ε4 and TREM2 have been identified as genetic risk factors for AD, and TREM2 has been shown to also bind anionic lipids, it has been hypothesised that TREM2 functions along similar mechanisms as APOE (9, 58). TREM2 is able to bind to APOE and APOE has been reported to be upregulated in microglia that surround amyloid plaques. The microglia that surround the neuritic plaques in AD were found to have a distinct phenotype to other microglial cells and have been termed “dark microglia,” disease‐associated microglia (DAM) or microglia of neurodegeneration (MGnD) by different groups (3, 26, 29). These microglia have been found to act differently to other activated microglia through the upregulation of several genes that stimulate the TREM2 pathway, including APOE (26).
Investigations have been undertaken on postmortem human brain samples (29, 46) carrying various TREM2 variants that have confirmed findings shown in mouse models. These included a reduction in plaque‐associated microglia, with no overall reduction in the number of microglia (29, 46), and an increase in neuritic plaque burden and overall tau burden in the hippocampus of TREM2 variant cases (46). Here we extend the investigations in human postmortem brains of sporadic AD cases with a TREM2 variant (TREM2+ SAD), sporadic AD cases without a TREM2 variant (TREM2− SAD), familial AD cases (FAD), normal controls with a TREM2 variant (TREM2+ controls) and normal controls without a TREM2 variant (controls). We hypothesised that TREM2+ cases would have a loss of function, indicating an altered inflammatory profile compared to TREM2− cases. Through immunohistochemical techniques, nanostring technology and label‐free quantitative mass spectrometry we show that whilst Aβ plaques and neurofibrillary tau tangles (NFT’s) do not significantly change between groups, the microglial phenotype is altered alongside differences in genetic and proteomic profiles between the TREM2− SAD and TREM2+ SAD cases and between the TREM2+ SAD and TREM2+ controls.
Material and Methods
Cases
All cases were obtained through the brain donation program at Queen Square Brain Bank for Neurological Disorders (QSBB). Standard diagnostic criteria were used to assess AD pathology in all cases (5, 39, 55). TREM2+ SAD cases (n = 3, 2 R47H, 1 D87N), TREM2+ control cases (n = 2, 2 R47H), TREM2− SAD cases (n = 19), FAD cases (n = 11) and control cases (n = 6) were used for this study. The demographic data for all cases is shown in Table 1. Cases were matched for age of onset and age at death where possible. There was no significant difference in age of onset between TREM2+ SAD cases and TREM2− SAD cases (P = 0.6769) or in age at death between TREM2+ SAD cases and TREM2− SAD cases (P > 0.9999). All TREM2 variant cases used in this study were genotyped as part of the study discovering TREM2 as a risk variant (16). Ethical approval for the study was obtained from the Local Research Ethics Committee of the National Hospital for Neurology and Neurosurgery.
| Case | Gender | Age of onset | Age at death | Disease duration | Postmortem delay (h) | Mutations | ApoE status | Clinical diagnosis | Pathological diagnosis | Braak and Braak | Thal | CERAD | ABC | ||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| TREM2SAD cases+ | |||||||||||||||
| 1 | M | 55 | 64 | 9 | 35:40:00 | R47H | E3/E4 | CBS | AD | 6 | 5 | Frequent | A3B3C3 | ||
| 2 | F | 56 | 66 | 15 | 51:20:00 | R47H | E4/E4 | SAD | AD | 6 | 5 | Frequent | A3B3C3 | ||
| 3 | M | 60 | 71 | 11 | 52:30:00 | D87N | E3/E4 | FTD | AD | 6 | 5 | Frequent | A3B3C3 | ||
| 4 | F | 36 | 41 | 5 | 64:15:00 | R47H Pre 200 PS1 | E3/E3 | SAD | AD | 6 | 5 | Frequent | A3B3C3 | ||
| TREM2control cases+ | |||||||||||||||
| 5 | M | – | 76 | – | 60:35:00 | R47H | E2/E2 | Control | Control | 0 | 0 | None | A0B0C0 | ||
| 6 | M | – | 82 | – | 25:30:00 | R47H | E3/E3 | Control | Control | 0 | 0 | None | A0B0C0 | ||
| TREM2SAD cases− | |||||||||||||||
| 7 | M | 63 | 73 | 10 | 31:10:00 | – | – | SAD | AD | 6 | 5 | Frequent | A3B3C3 | ||
| 8 | F | 51 | 63 | 12 | 16:00:00 | – | E3/E4 | SAD | AD | 6 | 5 | Frequent | A3B3C3 | ||
| 9 | F | 51 | 62 | 11 | 62:55:00 | – | E3/E4 | SAD | AD | 6 | 5 | Frequent | A3B3C3 | ||
| 10 | F | 65 | 70 | 5 | 46:58:00 | – | E3/E3 | SAD | AD | 5 | 5 | Moderate | A3B3C3 | ||
| 11 | M | 64 | 77 | 13 | 90:05:00 | – | E4/E4 | SAD | AD | 6 | 5 | Frequent | A3B3C3 | ||
| 12 | F | 49 | 62 | 13 | 76:40:00 | – | E3/E3 | SAD | AD | 6 | 5 | Frequent | A3B3C3 | ||
| 13 | M | 72 | 88 | 16 | 85:35:00 | – | E3/E4 | SAD | AD | 6 | 5 | Frequent | A3B3C3 | ||
| 14 | M | 52 | 69 | 17 | 35:04:00 | – | E3/E3 | SAD | AD | 6 | 5 | Frequent | A3B3C3 | ||
| 15 | M | 65 | 72 | 7 | 38:55:00 | – | E3/E4 | SAD | AD | 5 | 5 | Moderate | A3B3C3 | ||
| 16 | F | 76 | 85 | 9 | 90:20:00 | – | E3/E4 | SAD | AD | 6 | 5 | Frequent | A3B3C3 | ||
| 17 | M | 55 | 64 | 9 | 76:45:00 | – | E3/E4 | SAD | AD | 6 | 5 | Frequent | A3B3C3 | ||
| 18 | F | 69 | 74 | 5 | 93:40:00 | – | – | SAD | AD | 6 | 5 | Frequent | A3B3C3 | ||
| 19 | M | 80 | 85 | 5 | 129:15:00 | – | – | SAD | AD | 5 | 5 | Moderate | A3B3C3 | ||
| 20 | F | 46 | 52 | 6 | 51:55:00 | – | – | LBD | AD | 6 | 5 | Frequent | A3B3C3 | ||
| 21 | F | 49 | 55 | 6 | 47:50:00 | – | E3/E3 | SAD | AD | 6 | 5 | Frequent | A3B3C3 | ||
| 22 | M | 67 | 72 | 5 | 91:10:00 | – | E2/E4 | bvFTD | AD | 6 | 5 | Frequent | A3B3C3 | ||
| 23 | F | 65 | 79 | 14 | 22:30:00 | – | E3/E4 | SAD | AD | 6 | 5 | Frequent | A3B3C3 | ||
| 24 | M | 52 | 68 | 16 | 35:20:00 | – | E3/E4 | FTD/Picks | AD | 6 | 5 | Frequent | A3B3C3 | ||
| 25 | M | 58 | 68 | 10 | 52:05:00 | – | E3/E4 | SAD | AD | 6 | 5 | Frequent | A3B3C3 | ||
| FAD cases | |||||||||||||||
| 26 | F | 48 | 59 | 11 | 26:15:00 | PSEN1 202F | E4/E4 | FAD | AD | 6 | 5 | Frequent | A3B3C3 | ||
| 27 | F | 35 | 52 | 17 | 32:30:00 | PSEN1 Intron 4 | E4/E4 | FAD | AD | 6 | 5 | Frequent | A3B3C3 | ||
| 28 | M | 61 | 70 | 9 | 161:15:00 | PSEN1 S132A | E3/E4 | FAD | AD | 5 | 5 | Frequent | A3B3C3 | ||
| 29 | M | 42 | 51 | 9 | 43:10:00 | PSEN1 mutation | E3/E3 | FAD | AD | 6 | 5 | Frequent | A3B3C3 | ||
| 30 | F | 48 | 59 | 11 | 89:42:00 | V717L APP | E3/E3 | FAD | AD | 6 | 5 | Frequent | A3B3C3 | ||
| 31 | M | 60 | 66 | 6 | 68:05:00 | V717L APP | E3/E3 | FAD | AD | 6 | 5 | Frequent | A3B3C3 | ||
| 32 | M | 42 | 47 | 5 | 43:50:00 | PSEN1 A434T & T291A | E3/E3 | MSA | AD | 5 | 5 | Frequent | A3B3C3 | ||
| 33 | F | 46 | 66 | 20 | 31:55:00 | R278I | E3/E4 | FAD | AD | 6 | 5 | Frequent | A3B3C3 | ||
| 34 | F | 33 | 37 | 4 | 24:15:00 | E120K exon 5 PSEN1 | E3/E3 | FAD | AD | 6 | 5 | Frequent | A3B3C3 | ||
| 35 | F | 44 | 56 | 12 | 16:25:00 | APP V717I | E3/E3 | FAD | AD | 6 | 5 | Frequent | A3B3C3 | ||
| 36 | F | 39 | 47 | 8 | – | PSEN1 Intron 4 | E3/E3 | FAD | AD | 6 | 5 | Frequent | A3B3C3 | ||
| Control cases | |||||||||||||||
| 37 | M | – | 87 | – | 57:00:00 | – | E3/E3 | Normal | Normal | 2 | 3 | Moderate | A2B1C2 | ||
| 38 | M | – | 81 | – | 50:55:00 | – | E2/E2 | Normal | Normal | 2 | 1 | Sparse | A1B1C0 | ||
| 39 | F | – | 73 | – | 24:00:00 | – | E3/E4 | Normal | Normal | 2 | 2 | Sparse | A1B1C2 | ||
| 40 | M | – | 88 | – | 16:15:00 | – | E3/E3 | Normal | Normal | 2 | 2 | Moderate | A1B1C2 | ||
| 41 | F | – | 80 | – | 49:10:00 | – | E3/E3 | Normal | Normal | 0 | 2 | None | A0B1C0 | ||
| 42 | F | – | 93 | – | 29:40:00 | – | E3/E3 | Normal | Normal | 3 | 1 | Moderate | A1B2C2 | ||
| Averages | |||||||||||||||
| TREM2 | 2F:4M | 52 | 67 | 10 | 50:35:00 | – | – | – | – | 4 | 3 | – | A2B2C2 | ||
| SAD | 9F:10M | 60 | 70 | 10 | 61:48:00 | – | – | – | – | 6 | 5 | – | A3B3C3 | ||
| FAD | 7F:4M | 45 | 55 | 10 | 53:44:12 | – | – | – | – | 6 | 5 | – | A3B3C3 | ||
| Control | 3F:3M | – | 84 | – | 37:50:00 | – | – | – | – | 2 | 2 | – | A1B1C1 | ||
APOE genotyping
DNA extraction
About 100 mg of frozen cerebellum from all cases, except cases 7, 18‐20 where no frozen tissue was available was homogenised in extraction buffer (0.1 M NaCl, 20 mM Trizma base, 25 mM EDTA disodium, 0.5% SDS) and proteinase‐K solution (10 mg/mL) and samples digested at 55°C. A 1:1:1 mix of phenol, chloroform and IAA added and centrifuged for 5 minutes at 12 000 rpm. The aqueous layer was removed and 3M NaAC pH5.3 added. 100% ethanol was added to precipitate the DNA. The pellets were dried at room temperature and resuspended in TE (Tris‐EDTA) solution and stored at 4°C.
Genotyping
The Qiagen PCR Mix–GC Rich kit was used. The primers used to determine the ApoE status were previously reported (10). The master mix was added to the DNA and a PCR run with the following settings: 94°C for 5 minutes, 30× (94°C for 30 s, 60°C for 30 s, 72°C for 30 s), 72°C for 5 minutes before being left at 4°C. A 3% metaphor agarose gel/ 2% normal agarose was prepared with GelRed dye added. The digested PCR end product for each sample was added to the gel, which was run for 1 h 30 minutes at 80 V before being visualised in a DNR Bio‐Imaging Systems MiniBIS Pro.
Immunohistochemistry
Eight‐micron‐thick formalin‐fixed paraffin‐embedded (FFPE) tissue sections from the frontal cortex, temporal cortex and hippocampus were cut from the cases listed in Table 1. Sections were deparaffinised in xylene and rehydrated using graded alcohols. Immunohistochemistry for all antibodies required pressure cooker pre‐treatment for 10 minutes in citrate buffer pH 6.0. Aβ immunohistochemistry also required formic acid pre‐treatment prior to pressure cooking. Endogenous peroxidase activity was blocked in 0.3% H2O2 in methanol for 10 minutes and non‐specific binding blocked with 10% dried milk solution. Tissue sections were incubated with primary antibodies; Aβ (1:100; Dako); AT8 (tau, 1:600; Thermo); Iba1 (microglial, 1:1000; Wako); CD68 (microglial, 1:100, Dako); CR3‐43 (microglial, 1:150, Dako); P2RY12 (microglial, 1:100; Sigma); Glial fibrillary acidic protein (GFAP) (astrocytic, 1:1000 Dako) for 1 h at RT, followed by biotinylated anti‐rabbit IgG (1:200; Dako) or biotinylated anti‐mouse IgG (1:200; Dako) for 30 minutes at RT and Avidin‐Biotin complex (30 minutes; Dako). Colour was developed with di‐aminobenzidine/H202 (30). Stained sections were digitised using a Leica SCN400F slide scanner.
Double immunohistochemistry and Thioflavin‐S staining and analysis
Formalin‐fixed paraffin‐embedded tissue sections were cut from the frontal cortex of three TREM2+ SAD cases (cases 1‐3) and three TREM2− SAD cases (cases 13,15,17). Sections underwent immunohistochemistry as described above for: ionized calcium binding adapter molecule 1 (Iba1), CD68 and CR3‐43 using Tyramide Signal Amplification kit as the chromogen. Thioflavin‐S was applied for 7 minutes and differentiated with 70% ethanol. Slides were visualised under a Leica DM5500 fluorescent microscope. Z‐stack images at 63× magnification were collected from the frontal grey matter and antibody stain to analyse the microglial load observed in diffuse and dense core amyloid plaques. Using Image J, the channels were separated, the total plaque area calculated together with the percentage area positive for microglia markers. The sum of percentages for all 20 diffuse and all 20 dense core plaques were then taken for each case.
Pathological analysis
Digital morphological analysis
Digital images for Aβ, tau and microglial markers immunohistochemistry were viewed using Aperio Imagescope (v12.3.0.5056). We analysed the areal fraction of the immunohistochemical staining (all antibodies), as well as the number of stained microglia and the circularity of the microglia. Using Image J software (https://imagej.nih.gov/ij/↗) and a python script 10 randomised snapshots, representing 500 µm2, were generated from the extracted regions of interest. The region of interest included all six cortical layers of the cortex and all areas of the hippocampus. Bland‐Altman plots were performed to determine the reliability of the method and how many snapshots were needed. The 10 snapshots were then used to determine the areal fraction for each immunohistochemical preparation. Areal fractions were calculated for each snapshot and means were taken for each case, each region and each antibody. In the microglial stained preparations the number and the circularity of microglia were also determined. A score closer to one indicated the microglia were more amoeboid in shape whereas a score closer to 0 indicates the microglia were more ramified. An average circularity value was taken for each snapshot and a mean of the snapshots were taken for each case, region and stain. Kruskal‐Wallis one‐way ANOVA tests with Dunn’s multiple comparisons were performed to determine any significant differences for all types of analysis at a level of P < 0.05.
Manual analysis
The digital morphological analysis would not allow distinction between different Aβ plaque types, this analysis required manual counting. Using the randomized snapshots generated from the digital morphological analysis, different plaque types (diffuse or dense‐core) were manually counted to determine how many plaques were in the representative sample. Dense core plaques were only counted when a dense core could be visualised, otherwise they were counted as diffuse plaques.
Expression analysis
RNA extraction
RNA extraction was performed using the Qiagen RNeasy kit and protocol from 100 mg of frozen frontal cortex following the manufacturer’s instructions. The RNA concentration and purity was measured using an Eppendorf spectrophotometer.
Nanostring
All RNA samples were analysed on the NanoString Human Inflammation panel containing 256 genes and 30 extra genes relevant to AD. Excess probes were removed and probe/target complexes aligned and immobilised in the nCounter cartridges. Raw data was analysed on nSolver software (15). The results were normalised using positive controls and five housekeeping genes (CLTC, GAPDH, GUSB, PGK1 and TUBB). All pairwise ratios between groups were made from the normalised data and two‐tailed t‐tests were performed to establish any significance at P < 0.05.
Mass spectrometry
Frozen frontal cortex samples were homogenised in 50 mM Ambic buffer with 2% ASB‐14 and pooled per disease group (three cases per pooled sample other than TREM2+ controls that only had 2). Proteins were extracted into two fractions; soluble supernatant and the insoluble pellet fraction. For each fraction Label‐free mass spectrometry was performed with a SYNAPT G2‐Si High Definition mass spectrometer (Waters, UK) with 2D fractionation as previously described (8, 40). There were four fractions run for each sample and 0.5 µg of protein were injected per fraction per run. The raw data were imported into Progenesis for proteomics software (Nonlinear dynamics, UK) and processed. Identifications were obtained by searching the data against the human reference proteome (2016). Data for identifications with more than 1 unique peptide were exported for downstream analysis.
Bioinformatics
As a label‐free approach was taken, all genes or proteins that met the threshold set (P < 0.05 for nanostring, >1.5‐fold change compared to controls in expression for proteomics) were put into publicly available databases to assess the relationships between them and the biological processes, molecular functions and cell components that were enriched in the different disease groups. To assess the enriched gene ontology terms Webgestalt (62) was used. GOview was used to compare terms that were over‐represented between regions or disease group. Ingenuity Pathway Analysis software was used to perform in depth canonical pathway analysis.
Results
Pathological diagnosis ofcases TREM2 +
All cases underwent a routine diagnostic assessment. Routine immunohistochemistry was performed on each case to determine the final diagnosis (Table 1). Microscopic observations for all TREM2+ SAD cases (Table 1, cases 1‐3) showed cases reached end stage AD with a score of A3B3C3 according to current diagnostic criteria. Figure 1 demonstrating the presence of Aβ and tau in the hippocampus, cerebellum and occipital cortex, brain regions examined in the diagnostic criteria. Two cases (case 1 and 3) also had additional alpha‐synuclein pathology with Lewy bodies observed in the amygdala and substantia nigra. The two TREM2+ control cases (Table 1, cases 5 and 6) had no Aβ plaque pathology or tau pathology and were diagnosed as neurologically normal controls (Figure 1).
Aβ and tau immunohistochemistry inSAD,control andSAD cases. Aβ immunohistochemistry was carried out to confirm the diagnosis of AD. Aβ plaques were observed in the hippocampus and cerebellum in theSAD cases (and) andSAD cases (and). Whereas thecontrol cases showed no Aβ deposition (and). Tau immunohistochemistry showed severe neuropil threads, neurofibrillary tangles and neuritic plaques in the occipital cortex in theSAD cases () andSAD cases (). However, no tau positivity was observed for thecontrol cases (). Bar in a represents 100 µm in all panels. TREM2 + TREM2 + TREM2 − TREM2 + TREM2 − TREM2 + TREM2 + TREM2 − TREM2 + A D C F B E G I H
APOE status and pathological diagnosis
The APOE genotypes of cases used in this study are shown in Table 1. Table 2 shows the distribution of APOE alleles for all disease groups studied. There were predominantly more ɛ4 alleles present throughout the cases confirmed with AD pathology. The TREM2+ control cases did not carry any ɛ4 alleles. However, the TREM2+ SAD cases had at least one ɛ4 allele.
| Disease group | APOE genotype | |||||
|---|---|---|---|---|---|---|
| 2/2 | 2/3 | 2/4 | 3/3 | 3/4 | 4/4 | |
| Control | Y | Y | Y | |||
| TREM2SAD− | Y | Y | Y | Y | ||
| FAD | Y | Y | Y | |||
| TREM2Control+ | Y | Y | ||||
| TREM2SAD+ | Y | Y | ||||
No difference in amount and type of Aβ plaques inandSAD cases TREM2 + TREM2 −
Aβ load and types of plaques were determined in the different disease groups across three brain regions (Figure 2). There were no significant differences in Aβ load between TREM2− SAD and TREM2+ SAD cases (P = 0.6667). The FAD cases had significantly more Aβ in the hippocampus than both TREM2− SAD (hippocampus P = 0.0411). The Aβ load in the TREM2+ SAD cases mirrored the TREM2− SAD cases across all regions. Semi‐quantitative assessment was carried out to determine whether different plaque types were more or less prevalent in the TREM2+ SAD cases in the three brain regions. There was no significant difference in the numbers of dense cored plaques or diffuse plaques between the disease groups (TREM2+ SAD, TREM2− SAD and FAD) for all regions (Figure 2).
Quantitative analysis of Aβ and tau load and semi‐quantitative analysis of Aβ plaque type. The Aβ load was measured in theSAD,SAD and FAD cases (). Load was measured as mean % area stained in the frontal cortex (FCTX), temporal cortex (TCTX) and hippocampus (HIPPO). The numbers of diffuse or dense‐core plaques were determined in the same cohorts, in the frontal cortex (), temporal cortex () and hippocampus (). The tau load was determined in the cohorts (). Load measured as mean % area stained. Kruskal‐Wallis one way ANOVA was performed for each region, Wilcoxon‐paired ranks test was performed to determine statistical differences between types of plaques for each disease group in each region. Significance is shown as *< 0.05. TREM2 + TREM2 − P A C D E B
No difference in tau load by immunohistochemical analysis inandSAD cases TREM2 + TREM2 −
The level of tau accumulation was quantified in the three brain regions in TREM2+ SAD cases compared to TREM2− SAD and FAD cases (Figure 2). There was no significant difference in tau load between TREM2+ SAD cases and TREM2− SAD cases (P = 0.8068). However, there was a significant increase in tau load in the frontal cortex of FAD cases compared to TREM2+ SAD cases (P = 0.0248).
Differences in microglia load and morphology inandcases TREM2 + TREM2 −
Microglial load was assessed and compared between TREM2− SAD, TREM2+ SAD, FAD, TREM+ control cases and neurologically normal controls, using four microglial markers across three brain regions (Figure 3A–D). There were no significant differences that could be seen when assessing Iba1, CR343 or P2RY12 load (Figure 3I,M,O). However, TREM2+ SAD cases appeared to have a markedly increased Iba1 load in the frontal cortex and decreased load in the temporal cortex although this did not reach significance. In TREM2+ SAD cases CR343 load appeared to follow the same trend as TREM2− SAD cases when compared to other disease groups. However, the TREM2+ control cases had significantly less CR343 positive, activated microglia, present compared to the TREM2+ SAD cases. P2RY12 load remained consistent between disease groups although a greater level of variation was observed in the hippocampus compared to other regions, with TREM2+ control cases having markedly less homeostatic microglia than the TREM2+ SAD cases. TREM2+ SAD cases had a significantly higher CD68 load in the frontal cortex compared to the TREM2+ control cases (Figure 3K; P = 0.0147) and the FAD cases (Figure 3K; P = 0.0127). Representative CD68 images from TREM2− SAD, TREM2+SAD, FAD and TREM2+ control cases illustrate these differences (Figure 3E–H).
As microglia have a spectrum of different morphological phenotypes, ranging from a ramified, surveillance phenotype to an amoeboid phagocytic phenotype, the shape of the microglia were also investigated across the different disease groups. This was measured as an index of circularity in which a perfect circle (amoeboid microglia) scored closer to 1 and an imperfect shape (ramified microglia) scored closer to 0. TREM2+ SAD cases had significantly more amoeboid Iba1 microglia than controls in the frontal and temporal cortices (Figure 3J; P = 0.0248). TREM2+ SAD cases also had significantly more circular CR343 microglia than the TREM2− SAD cases (Figure 3N; P = 0.0128) in the frontal cortex and more circular microglia than the controls (P = 0.0064) in the temporal cortex. Furthermore, TREM2+ SAD cases had significantly more circular CD68 microglia than the controls in the frontal cortex (Figure 3L; P = 0.0296). As expected P2RY12 microglia generally had a circularity score closer to 0 than Iba1, CD68 and CR3‐43 microglia (Figure 3P). No major differences were observed between disease groups for P2RY12 circularity.
Microglial load and circularity inandcases. Measured using the microglial markers Iba1, CD68, CR3‐43 and P2RY12 in frontal cortex (FCTX), temporal cortex (TCTX) and hippocampus (HIPPO) betweenSAD,SAD, FAD,controls and normal controls. Representative images of microglial immunohistochemistry for Iba1 (), CD68 (), CR3‐43 () and P2RY12 () in case 7. Representative images of CD68 staining in the frontal cortex of aSAD (, case 13), FAD (, case 27),SAD (, case 2) andcontrol (, case 3). Iba1 load (), CR343 load (), CD68 load (), and P2RY12 load (). A measure of circularity was obtained where 1 was a perfect circle whereas a score of 0 was an imperfect shape. Mean values were taken from each case in a disease group. The circularity scores are shown Iba1 circularity (), CD68 circularity (), CR3‐43 circularity (), and P2RY12 circularity (). Black arrows highlight CD68 + ve microglia. Scale bar represents 50 µm. Significance is shown as *< 0.05, **< 0.005. TREM2 + TREM2 − TREM2 + TREM2 − TREM2 + TREM2 − TREM2 + TREM2 + P P A B C D E F G H I J K L J L N P
No differences in microglial clustering around amyloid plaques inandSAD cases TREM2 + TREM2 −
The percentage area of microglial staining around diffuse and cored plaques was analysed in TREM2− and TREM2+ SAD cases to determine if the presence of the TREM2 variant correlated with the amount of microglial positive staining that clustered around amyloid plaques. Three microglial markers (Iba1, CD68 and CR343) were used to determine differences between different activation states of microglia around the plaques. There were no significant differences between TREM2+ SAD cases and TREM2− SAD cases in any of the three markers or between diffuse and dense core amyloid plaques (Figure 4). A trend was observed showing TREM2+ SAD cases had less percentage area load of CD68 in both diffuse and dense core plaques than SAD cases; however, this did not reach significance.
Clustering of microglia around amyloid plaques. Three microglial markers [Iba1 (), CD68 () and CR3‐43 ()] were used to investigate the area of microglial staining around either dense cored plaques or diffuse plaques. Representative images of dense cored plaques and diffuse plaques with double staining with thioflavin (green) and microglia markers (red). No significant differences were found in the area of microglial staining betweenSAD andSAD cases. Bar on image represents 25 µm. A B C TREM2 − TREM2 +
variant effect on astrocyte phenotype TREM2
GFAP is a reactive astrocyte marker. GFAP was found to be upregulated to a higher level in TREM2+ SAD cases at both the gene and protein level. At the gene expression level GFAP was upregulated 1.99‐fold in TREM2− SAD cases, 2.35‐fold in TREM2+ SAD cases and 1.3‐fold in TREM2+ control cases all compared to neurologically normal controls. At the protein level GFAP was upregulated across all groups but the greatest level of upregulation was observed in the TREM2+ SAD cases in both the soluble and insoluble fractions (3.8‐fold soluble, 3.2‐fold insoluble). As this increase in expression was observed for the TREM2+ SAD cases, we assessed the GFAP pathological load in the frontal cortex, temporal cortex and hippocampal regions (Figure 5). There were no significant differences between TREM2+ SAD cases, TREM2− SAD cases and TREM2+ control cases. However, there were significant differences between FAD cases and TREM2+ SAD cases.
Astrocyte load inandcases. Measured using GFAP in the frontal cortex (FCTX), temporal cortex (TCTX) and hippocampus (HIPPO) betweenSAD,SAD, FAD,controls and normal controls. No significant differences were observed betweenandcontrol cases. However, significant differences were observed between FAD andSAD,controls and normal controls. Significance is shown as *< 0.05. TREM2 + TREM2 − TREM2 + TREM2 − TREM2 + TREM2 + TREM2 − TREM2 + TREM2 + P
Expression analysis ofSAD andcontrols TREM2 + TREM2 +
The NanoString Technologies Human Inflammation panel containing 256 genes and 30 targeted genes specific to AD were processed and analysed on the frontal cortex from all groups and compared with normal controls. Upregulation or downregulation of genes are determined as a ratio to normal controls. A proportion of the genes analysed were upregulated in TREM2− SAD (124 genes at a significance level of P < 0.05, of which 70 genes at P < 0.01 and 24 genes at P < 0.001). Two genes were downregulated significantly in TREM2− SAD, MRC1 (macrophage mannose receptor 1) and SNCA (alpha‐synuclein) (P < 0.05); in FAD (91 genes were significantly upregulated at level of P < 0.05, 22 genes at P < 0.01 and 2 genes, NFKB1 (nuclear factor NF‐kappa‐B p105 subunit) and TGFB2 (transforming growth factor beta‐2) at P < 0.001). Two genes were downregulated PIK3C2G (Phosphatidylinositol 4‐phosphate 3‐kinase C2 domain‐containing subunit gamma) and SNCA (P < 0.05); in TREM2+ SAD cases (22 genes were significantly upregulated at a level of P < 0.05); in TREM2+ controls (11 genes were significantly upregulated and 14 genes were downregulated P < 0.05) (Supporting Table S1). Visualisation of the canonical pathways and biological functions altered between the disease groups and normal controls highlights the number of pathways predicted to be activated in TREM2+ SAD, TREM2− SAD and FAD. Moreover, the majority of pathways represented were predicted to be inhibited in the TREM2+ controls cases (Supporting Figure S1). There were several genes that were significantly altered across all disease groups. Genes that demonstrated more than twofold change in expression compared to controls across all three groups included HSPB2 (Heat shock protein beta‐2), TGFB1 (Transforming growth factor beta‐1), CSF1 (Macrophage colony‐stimulating factor 1) and CXCR4 (C‐X‐C chemokine receptor type 4). Several genes showed significantly altered expression in only one disease group. For example, APOE and PEN2 (Gamma‐secretase subunit PEN‐2) were only significantly upregulated in TREM2− SAD compared to controls but were not significantly altered in any other disease group. Similarly, CD68, IL6R (Interleukin‐6 receptor subunit alpha) and TYROBP (TYRO protein tyrosine kinase‐binding protein) were significantly upregulated in FAD compared to controls. TREM2+ SAD cases were the only group to have a significant upregulation of GRB2 (Growth factor receptor‐bound protein 2).
Proteomic analysis ofSAD andcontrols TREM2 + TREM2 +
To determine if the genetic expression changes also occur at the translational level, label‐free mass spectrometry was performed to determine changes in protein expression. A total of 6012 proteins were detected in the soluble supernatant fraction and 5571 in the insoluble pellet fraction. Proteins were either detected in both the soluble and insoluble fraction (3269 proteins) or they were unique to the soluble fraction (2743 proteins) or insoluble fraction (2302 proteins). Only proteins that were changing >1.5‐fold compared to normal controls were included (3330 proteins) and both up‐ and downregulation of proteins were observed in every disease group when compared to controls. The top 20 upregulated and downregulated proteins are listed in Table 3. There was very little overlap in these proteins between the different disease groups.
Ingenuity pathway analysis highlighted that the top canonical pathways (predicted to be activated or inhibited by the expression of proteins represented in them) differ between the soluble and insoluble fractions. The top pathways, represented by the identified proteins in the soluble fraction (Supporting Figure S2) are HIPPO signalling, Melatonin signalling, LXR/RXR activation, amyloid processing, actin cytoskeleton signalling and neuroinflammation signalling. HIPPO signalling, amyloid processing and neuroinflammation signalling are predicted predominantly to be activated across the disease groups, whereas melatonin signalling, LXR/RXR activation and actin cytoskeleton signalling predicted activation and inhibition were more mixed between groups (Supporting Figure S2). The top canonical pathways in the insoluble fraction (Supporting Figure S2) are EIF2 signalling, RhoA signalling, signalling by Rho family GTPases, ERK5 signalling, glioma signalling and ephrin receptor signalling. EIF2 signalling and glioma signalling are predominantly predicted to be inhibited across the disease groups with TREM2+ control cases having the opposite prediction. RhoA signalling, signalling by Rho family GTPases, ERK5 signalling and ephrin receptor signalling have a more activated prediction. The TREM2+ control group has a different pattern showing predominantly predicted activation throughout the top canonical pathways, whereas the other groups are more varied between activation and inhibition.
| SADTREM2− | FAD | SADTREM2+ | ControlTREM2+ | ||||
|---|---|---|---|---|---|---|---|
| Upregulated | |||||||
| KRT6A | 25.32 | STOX2 | 281.43 | GOLGA8R | 60.79 | PSMB7 | 13.12 |
| KRT6C | 8.93 | SNRPD3 | 219.66 | APOE | 7.77 | NEDD8 | 9.55 |
| DSC1 | 8.47 | TMEM43 | 152.18 | MAPT | 7.74 | WDR19 | 9.32 |
| COL25A1 | 7.94 | QRICH2 | 136.53 | ECT2 | 7.14 | SLC14A1 | 5.92 |
| KRT5 | 6.57 | FUS | 131.6 | IFIT3 | 7.11 | C2CD3 | 5.37 |
| KRT16 | 6.5 | CUL5 | 108.19 | C4A | 6.83 | SHTN1 | 4.89 |
| CRTC1 | 6.07 | TRPM1 | 86.62 | FHAD1 | 6.27 | PGBD5 | 4.13 |
| APOE | 6.01 | PEX19 | 84.41 | SPG7 | 6.21 | WDR47 | 4.05 |
| KRT17 | 5.4 | SETD5 | 79.52 | PHF5A | 5.04 | IGHA2 | 3.88 |
| MAPT | 4.49 | CDC42BPG | 78.53 | CPSF7 | 4.8 | LAMA5 | 3.81 |
| C4A | 4.29 | PITHD1 | 73.86 | RIC8A | 4.58 | SCYL1 | 3.73 |
| IRF2BP1 | 3.78 | RAP1GAP | 68.06 | ANXA2 | 4.41 | MAP2K1 | 3.63 |
| OBSCN | 3.37 | WASL | 67.96 | STXBP6 | 4.41 | SLC25A5 | 3.61 |
| QARS | 3.35 | GLO1 | 66.47 | HIST1H4A | 4.23 | PSAP | 3.5 |
| COL9A2 | 3.29 | DNM1L | 65.49 | ZBTB38 | 4.21 | NEB | 3.5 |
| KRT1 | 3.13 | CUX2 | 65.09 | SQSTM1 | 4.16 | PARP1 | 3.29 |
| ASAP2 | 3.1 | ATPAF1 | 56.8 | GRIA2 | 4.09 | KRT10 | 3.27 |
| UBN1 | 3.08 | RHOA | 55.35 | RBM25 | 3.98 | LAMA2 | 3.25 |
| ARFGEF1 | 3.03 | GRB2 | 54.74 | APP | 3.91 | CCDC30 | 3.16 |
| DCTN2 | 2.96 | UTP15 | 54.63 | PSMC4 | 3.85 | MAP1B | 3.14 |
| Downregulated | |||||||
| ASAP3 | −7.45 | TTC23 | −72.67 | CCNG1 | −18.89 | ASAP3 | −54.57 |
| VGF | −4.32 | INPP4A | −61.52 | ASAP3 | −12.19 | CDH8 | −48.09 |
| TBC1D8B | −4.16 | SEC14L2 | −46.27 | GPATCH2 | −8.72 | WDFY1 | −11.28 |
| SCG2 | −3.16 | TCEB2 | −15.76 | KSR1 | −4.69 | COX6A1 | −9.54 |
| USP4 | −2.91 | APCS | −9.98 | EPB42 | −4.30 | SRRM2 | −8.96 |
| VPS51 | −2.80 | TNIK | −8.80 | ZEB2 | −4.22 | AGFG1 | −7.18 |
| HIST1H1E | −2.55 | PGAM5 | −7.15 | SRRM2 | −3.66 | RAB24 | −6.49 |
| KPNA6 | −2.52 | KIAA1468 | −7.01 | FBF1 | −3.53 | MKRN1 | −6.47 |
| TCEAL3 | −2.48 | OGDHL | −6.51 | ASNS | −3.45 | PXN | −6.28 |
| TPM2 | −2.47 | KTI12 | −5.07 | SI | −3.27 | PANK4 | −6.08 |
| FARSA | −2.44 | SAMD11 | −4.15 | CCDC51 | −3.10 | GC | −5.48 |
| SYNPO | −2.43 | LYN | −4.07 | TAGLN | −2.89 | H3F3A | −5.42 |
| ALDH3A2 | −2.40 | SDSL | −4.03 | NUCKS1 | −2.84 | HBB | −4.95 |
| IRF2BP2 | −2.38 | PIP4K2A | −4.02 | CIRBP | −2.76 | FDFT1 | −4.82 |
| TNRC18 | −2.35 | NPM1 | −3.95 | HNRNPUL1 | −2.69 | TECPR1 | −4.75 |
| TAGLN | −2.31 | HNRNPLL | −3.89 | PXN | −2.67 | CA1 | −4.52 |
| PRIM1 | −2.30 | HNRNPM | −3.79 | C2CD3 | −2.65 | PTPRE | −4.50 |
| KSR1 | −2.29 | KLC2 | −3.59 | EPC2 | −2.64 | CUX2 | −4.43 |
| SENP6 | −2.21 | IFIT3 | −3.59 | PITPNM3 | −2.60 | ERO1A | −4.40 |
| SLC12A7 | −2.20 | SNRPD1 | −3.58 | WNK2 | −2.60 | VARS | −4.28 |
Proteomic expression of known AD markers
Several proteins known to be related to AD or neurodegeneration were also altered (Table 3). APP and MAPT are the proteins directly related to AD pathology and in both cases these were upregulated in all groups other than the TREM2+ control cases, in which they were downregulated when compared to normal controls. For APP, 3.9‐fold change was observed in the TREM2+ SAD cases in which it was one of the highest 20 proteins observed to be upregulated. For MAPT, the greatest fold change difference was also observed in the TREM2+ SAD cases (7.7‐fold). It was in the highest 20 proteins upregulated for TREM2− SAD and TREM2+ SAD. GFAP was upregulated across all groups but the greatest level of upregulation was observed in the TREM2+ SAD group in both the soluble and insoluble fractions (3.8‐fold soluble, 3.2‐fold insoluble). APOE was upregulated across all groups in the insoluble fraction, being in the highest 20 proteins upregulated for all groups other than TREM2+ control cases. Again, the highest level of upregulation was observed in the TREM2+ SAD cases (7.8‐fold, insoluble fraction).
Multi‐omic comparison
The genetic expression data obtained through nanostring methods and the protein expression data using label‐free mass spectrometry were compared with its limitations in mind. As the nanostring data only investigated 256 neuroinflammatory genes and 30 genes specific to neurodegeneration, neuroinflammatory pathways are more likely to appear as represented when comparing the two datasets. It is, therefore, no surprise that when the top canonical pathways represented in each disease group and using each experiment were compared, the neuroinflammatory signalling pathway was one of the top pathways represented across disease groups. (Supporting Figure S3). However, as this pathway is one of the top canonical pathways represented in the proteomics data, it shows that neuroinflammation is still a major factor at the protein level and should be investigated thoroughly. The neuroinflammation signalling pathway is predicted to be activated in both the nanostring and proteomic data of all groups apart from the TREM2+ control group, in which it is inhibited. When comparing the canonical pathways, the general pattern shows that the pathways are predicted to be activated in the genetic expression data but inhibited in many of the proteomic data. Additionally, the TREM2+ control group had a different profile in that the pathways most highly represented in this group are predicted to be inhibited at the genetic level but activated at the protein expression level. This again highlights how different the TREM2+ control group are to not only the other AD groups but the other cases that have a TREM2 variant. This suggests that there is regulation between the transcriptional and translational level that needs further investigation.
Discussion
We undertook a morphological and biochemical study on a cohort of SAD and control cases found with and without a TREM2 variant and FAD cases. All TREM2+ SAD cases were diagnosed with AD and all carried an APOE ε4 allele. The TREM2+ control cases, were negative for an APOE ε4 allele and had no Aβ plaque or tau pathology. When comparing the pathological burden between TREM2+ and TREM2− SAD cases we found no differences in the Aβ plaques types, overall Aβ load or tau burden in three different brain regions. Our results are in support of recent studies showing there is no difference in the overall levels of insoluble Aβ between TREM2+ and TREM2− SAD cases (46) and no difference in the composition of the Aβ plaque types. From an Aβ perceptive, the TREM2+ SAD cases are indistinguishable from TREM2− SAD cases. Whereas, in mouse models overexpressing TREM2, decreases in amyloid plaque deposition were observed, and TREM2 deficient models showed conflicting results on the amount of amyloid load between cortical and hippocampal regions (20, 21, 58, 59). This data agree with recent findings that TREM2 acts differently in mice and humans (60, 63). We found no significant differences in overall tau load between TREM2+ and TREM2− SAD cases in three different brain regions, which is in contrast to a recent postmortem study (46). It has also been shown in TREM2 deficient mouse models that tau was increased compared to wild types (22, 23).
We used four markers to investigate the microglial load and microglial morphology, with no difference in the microglial load observed with CR343, Iba1 or P2RY12. Iba1 was used as a pan microglial marker that detects ramified and amoeboid forms of microglia (4, 53). CR343, is a glycoprotein that is part of the major histocompatibility complex class II (MHC II) subgroup and was used to identify reactive/activated microglia (11, 36). P2RY12 was identified as a gene unique to microglia and represents homeostatic microglia (6, 17). This is in keeping with a recent postmortem investigation that showed no difference in the microglial load between high AD and TREM2 variant cases (46). However, an increase in CD68 positive microglia was seen in the frontal cortex in the TREM2+ SAD cases compared to both the TREM2+ controls and controls. CD68 was used as a phagocytic marker, as it is a lysosomal protein found within microglia, monocytes and macrophages (4). Although we found no difference in Iba1 and CR343 load between cases, we observed that the Iba1 positive microglia were more amoeboid in the TREM2+ SAD cases than controls in the frontal and temporal cortices. Whereas the CR343 positive microglia were more amoeboid in the TREM2+ SAD cases than the TREM2− SAD and FAD cases. We investigated the spatial relationship of the microglia around both dense and diffuse plaques and found that with three microglia markers (Iba1, CD68 and CR343) there were no differences in the amount of microglial staining around the plaques in any of the disease groups. However, the TREM2+ control cases had significantly less CR343 positive, activated microglia, present compared to the TREM2+ SAD cases. Microglia in AD are thought to be responsible for clearing Aβ plaques via phagocytic mechanisms and TREM2 is thought to have a role in these mechanisms (4, 16, 25). Our data would suggest that the microglia in the TREM2+ SAD cases are capable of reacting and forming amoeboid microglia. However, our results are not able to show whether the microglia are actually capable of phagocytosing the Aβ plaques. Mazaheri et al (2017) showed that TREM2 deficient microglia have reduced chemotaxis and response to neuronal injury. Alterations in the phagocytic mechanisms in these TREM2 variant microglia may play a part in their reduced response to neuronal injury.
To decipher the morphology and distribution of TREM2 within the brain and its relationship to microglia and Aβ plaques, we attempted to perform TREM2 immunohistochemistry. Seven different commercial TREM2 antibodies were tried in order to distinguish the level of pathological load of TREM2 in the TREM2+ variant cases compared to TREM2− cases (data not shown). To identify if these were specific we used double immunofluorescence with microglial markers and found no co‐localisation between TREM2 and the microglial markers. Therefore, we were not convinced at the specificity of the TREM2 antibodies. There have been some studies that report to show human TREM2 immunohistochemistry but with varying results (12, 51). Most recently Raha‐Chowdhury et al (47, 48) have shown some immunohistochemistry with TREM2 in human tissue but it is unknown how well this was characterised. There has been some debate in the field when some of these studies were published that TREM2 was only expressed in monocytes, hence the staining pattern separate to microglia (7, 12, 51). However, a number of high profile transcriptomic studies have now shown that TREM2 is expressed on DAM/neurodegenerative phenotype of microglia (3, 26, 29). Having said this, TREM2 immunohistochemistry has been shown in other models such as primary cells (19) and mouse models (7, 14). These studies show that TREM2 can be found within myeloid cells surrounding amyloid plaques.
We assessed the levels of expression of a number of genes involved in inflammation and known genes involved in AD pathogenesis. A striking difference was seen between the TREM2+ SAD, TREM2− SAD and FAD cases compared to the TREM2+ control cases. Visualisation of the canonical pathways and biological functions showed an almost opposite effect between the TREM2+ SAD cases and the TREM2+ control cases whilst, the TREM2+ controls cases showed a downregulation of the majority of the pathways represented. This included neuroinflammatory pathways, predicted to be activated in all AD groups and downregulation only observed in the TREM2+ controls cases. This suggests that TREM2 variants drive or fuel disease mechanisms once the disease has been initiated. We investigated the protein expression changes in the different groups compared to the controls. There was very little overlap in the top 20 upregulated and downregulated proteins. One overlap was the downregulation of Paxillin (PXN) in both the TREM2+ control cases and TREM2+ SAD cases. PXN is a cytoskeletal protein involved in actin‐membrane attachment at sites of cell adhesion to the extracellular matrix and involved in cell movement and migration (31). This suggests that the microglia cells in the TREM2+ cases are less mobile than TREM2− cases. This would need further investigation on a homogeneous microglial cell population extracted from postmortem human brain to confirm these findings are associated with the microglia.
We used ingenuity pathway analysis to highlight the top canonical pathways predicted to be activated or inhibited. The top pathways, represented by the proteins identified in the soluble fraction, activated across all disease groups included HIPPO signalling, amyloid processing and neuroinflammation, whereas melatonin signalling, LXR/RXR activation and actin cytoskeleton signalling predicted activation and inhibition were more mixed between groups. The top canonical pathways inhibited across disease groups included EIF2 signalling and glioma signalling. However, like the gene expression analysis the TREM2+ control cases, in the majority of pathways, had the opposite prediction. TREM2+ SAD cases had the highest levels of upregulation of APP protein, TREM2+ controls were the only group to show downregulation of APP. We found in both gene expression and proteomic investigations an increase in GFAP in the TREM2+ SAD cases. However, when we assessed the GFAP pathological load in three different brain regions there were no significant differences between the TREM2+ SAD cases, TREM2− SAD cases and TREM2+ control cases. This would suggest that different post‐translational modifications in GFAP could be implicated in these differences that are not detected using immunohistochemical techniques.
The main limitation of this study were the number of cases that were available to be studied. Although our results are validated by differences observed here replicating some findings observed in larger studies, none of these studies have looked at the impact TREM2 variants are having in cases that do not develop Alzheimer’s disease. The novelty of our study is the reporting of the TREM2+ control cases that were cognitively normal and did not have any amyloid plaques or tau pathology present. It has been reported that TREM2 expression decreases as we age and the brain becomes more vulnerable to neurodegeneration (37). The identification of similar TREM2+ control cases could potentially aid in understanding how TREM2 variants are involved in disease pathogenesis.
In summary we have shown that TREM2+ SAD cases, with a R47H or D87N variant, were pathologically indistinguishable from TREM2− SAD cases, when investigating the Aβ and tau loads. We have shown that TREM2+ control cases with a R47H variant, showed no Aβ or tau deposition. All TREM2+ SAD cases carried an APOE ε4 allele, whereas the TREM2+ control cases did not carry an APOE ε4 allele. We have previously reported an association between the TREM2 R47H variant and the APOE ε4 allele, which can also be seen throughout the cases identified with TREM2 R47H variant in the literature (28, 29, 61). Therefore, the hypothesis that an APOE ε4 genotype predisposes to the disease and a TREM2 R47H variant drives the pathogenesis of AD was introduced (41). Although in a recent paper only 50% of the TREM2 variant carriers were positive for at least one APOE ε4 allele, however the authors did not provide details on the specific TREM2 variant present (46). It will, therefore, be important to analyse the impact of APOE genotypes on specific TREM2 variants rather than combining the TREM2 variants into one group. To date there has been no confirmation that different APOE alleles bind to TREM2 with differing affinities. However, the TREM2 R47H variant has been shown to have reduced ligand binding, including lipoproteins (1). The observation that TREM2+ control cases do not have any AD pathology at time of death, have a strikingly different gene expression and proteomic profile compared to the TREM2+ SAD cases, TREM2− SAD cases and normal controls, suggests that other factors initiating the disease process are absent.
Supporting information
Acknowledgements
The authors have no conflict of interest. TL is supported by an Alzheimer’s Research UK senior fellowship. CT was supported by an Alzheimer’s Research UK PhD Studentship and is now supported by the Michael J Fox Foundation. BB was supported by the Leonard Wolfson Centre for Experimental Neurology. The Queen Square Brain Bank is supported by the Reta Lila Weston Institute for Neurological Studies and the Medical Research Council. This research was partly supported by the UK Dementia Research Institute at UCL and the National Institute for Health Research (NIHR) Queen Square Biomedical Research Unit in Dementia based at University College London Hospitals (UCLH), University College London (UCL). We thank the Peto Foundation for their kind donations to the UCL mass spectrometry research group. This work is (partly) funded by the NIHR GOSH BRC. The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR or the Department of Health. Alzheimer’s Society for funding an undergraduate summer student (GP). We also thank Dr Yau Lim for the initial development of the analysis macro, which has been tailored to the needs of this study.
Data Availability Statement
The data that support the findings of this study are available from the corresponding author upon reasonable request.
References
Associated Data
Supplementary Materials
Data Availability Statement
The data that support the findings of this study are available from the corresponding author upon reasonable request.