Rong Li1, Xia Wang2, Pengfei He1. 1. Department of Health Management, Yuncheng Central Hospital, Shanxi Medical University, Yuncheng, Shanxi 044000, P.R. China. 2. Drug Clinical Trial Center, Yuncheng Central Hospital, Shanxi Medical University, Yuncheng, Shanxi 044000, P.R. China.
Alzheimer's disease (AD), the most prevalent form of dementia in the elderly, is characterized by the progressive loss of cognitive function. It is pathologically characterized by the extracellular accumulation of amyloid β proteins and intracellular formation of neurofibrillary tangles (1). Accumulating evidence has suggested an important role of the genetic component, which is estimated to be 60-80%, in the pathogenesis of AD (2). Numerous susceptible loci containing common variants [minor allele frequency (MAF) >5%] have been identified in recent decades (3). However, only the Ɛ4 allele of apolipoprotein E (APOE) is confirmed to be the most important variant conferring a 3- to 4-fold risk to AD (4). Furthermore, since most of the loci reside in the intronic or intergenic regions, the biological mechanisms remain difficult to interpret. With the advent of next-generation sequencing, more and more rare coding variants (MAF<1%) with a moderate effect on the risk of AD, including those in phospholipase Cγ2 (PLCG2), ABI family member 3 (ABI3) and triggering receptor expressed on myeloid cells 2 (TREM2), have been identified (5).TREM2 is located on chromosome 6p21.1 and encodes a transmembrane protein on microglial cells. The protein is involved in the innate immunity within the central nervous system (CNS) by stimulating phagocytosis and inhibiting cytokine production (6). The association between TREM2 and AD was first suggested by findings indicating that homozygous loss-of-function variants of TREM2 caused autosomal recessive Nasu-Hakola disease, which is characterized by polycystic lipomembranous osteodysplasia with sclerosing leukoencephalopathy and early-onset dementia (7). Subsequent genome-wide and exome-wide sequencing revealed that rare coding variants of TREM2 may confer a risk of AD (4,8). A rare missense variant, p.Arg47His (R47H, rs75932628), has been identified as a risk factor of AD in Caucasians (4,8), but not in African-American or Asian populations (9,10). Additional rare variants, including p.Arg62His (R62H, rs143332484), p.Asp87Asn (D87N, rs142232675) and p.His157Tyr (H157Y, rs2234255), have also been detected in AD subjects and indicated to increase the susceptibility to AD (11,12). Furthermore, certain TREM2 mutations, such as p.Gln33stop, p.Tyr38Cys and Thr66Met, have been identified only in patients with AD but not in normal controls, and appear to be causative (13), However, due to the low frequency of these variants, studies with a small sample size may not have adequate power to identify the genetic associations (14-16). The results of association studies from different regions and ethnicities are not always consistent. Therefore, a meta-analysis for the genetic association between TREM2 rare coding variants and AD risk involving 73,128 individuals was performed.
Materials and methods
Search strategy
The PubMed, Embase and Web of Science databases were searched for articles investigating the association between TREM2 variants and the risk of AD using the following key words: (‘TREM2’ OR ‘triggering receptor expressed on myeloid cells 2’) and (‘AD’ OR ‘Alzheimer's disease’ OR ‘Alzheimer disease’) from inception to December 31, 2019. Additional articles were obtained by manually searching the reference lists of review and research articles.
Inclusion and exclusion criteria
For inclusion in the present meta-analysis, studies were required to meet the following criteria: i) Investigation of the association of TREM2 variants with the risk of AD in human subjects; ii) reporting on genotype data of at least one of the following four variants: R47H, R62H, D87N and H157Y; iii) studies with a case-control design with unrelated samples; and iv) providing sufficient data to calculate an odds ratio (OR) and corresponding 95% CI. If studies had overlapping samples, only the study with a larger sample size was included. Studies were excluded if they were based on family samples, used imputed data or provided insufficient genotype data.
Data extraction
A total of two independent researchers (RL and XW) extracted the following information: First author, publication year, ethnicity, sample size, genotype distribution in the case and control group and the genotyping method. Discrepancies between researchers were resolved by discussion.
Quality assessment
The present study used the Newcastle-Ottawa Scale (NOS) (17), which includes 8 items in 3 categories (selection, comparability and exposure), to assess the quality of all included studies. There are 9 stars in total, and studies having 7 or more stars were considered of high quality.
Statistical analysis
Since the four variants examined are rare in the general population, only the AD risk of carriers (heterozygous + homozygous variants) vs. non-carriers (homozygous wild-type) was compared. The OR was calculated as follows: OR=(no. of case carriers/no. of case non-carriers)/(no. of control carriers/no. of control non-carriers) and the 95% CI was determined to estimate the association strength of each variant with AD risk. Between-study heterogeneity was assessed by I2 statistics, which indicated low, intermediate and high heterogeneity if I2<25, 25-50 and >50%, respectively. The random-effects model was applied if there was high heterogeneity; otherwise, the fixed-effects model was used. Sensitivity analysis was performed to evaluate the impact of each study on the pooled effect size by excluding one study at a time. Publication bias was detected by Begg's funnel plots and Egger's test. All statistical analyses were performed by using STATA 12.0 (StataCorp LP). P<0.05 was considered to indicate statistical significance.
Results
Study selection
A total of 33 studies investigating the genetic association between rare TREM2 variants (R47H, R62H, D87N and H157Y) and the risk of AD were identified. However, 9 studies were excluded for the following reasons: A total of 5 did not provide any variant of interest in neither the case nor control groups (18-22), 3 used imputed data (5,23,24) and 1 did not provide genotype data (25). A total of two studies used a multi-stage strategy (8,26), and the datasets from discovery and replication stages were independently included in the quantitative synthesis. Finally, a total of 26 datasets from 24 studies (4,8-12,14-16,26-40) comprising 28,007 cases and 45,121 controls were eligible for the meta-analysis. Among these studies, 24 were for R47H (4,8-11,14-16,26-37,39,40), 12 for R62H (8,9,11,14-16,27,28,31,32,38,39), 11 for D87N (8,9,11,15,26-28,31,32,36) and eight for H157Y (8-12,26,31,36). A flowchart of the literature search is provided in Fig. 1. Regarding ethnicity, the majority of screened individuals were Caucasians (20 datasets) and the others were from African (2 datasets), Asian (2 datasets) or mixed populations (2 datasets). All studies were of high quality (≥7 stars) according to the NOS. The characteristics of all datasets included in the quantitative synthesis are listed in Table I.
Figure 1
Flowchart of the literature search. TREM2, triggering receptor expressed on myeloid cells 2.
Table I
Characteristics of all datasets included in the quantitative synthesis.
A total of 24 datasets (26,847 cases and 43,609 controls) investigated the association between the R47H variant and risk of AD. Carriers of R47H accounted for 1.39% of ADpatients and 0.35% of controls. There was no between-study heterogeneity (I2=0) and the fixed-effects model was used. Compared with non-carriers, R47H carriers were more susceptible to AD (OR=3.88, 95% CI: 3.17-4.76, P<0.001; Fig. 2). Subgroup analysis of the datasets of Caucasian populations (20 datasets with 22,175 cases and 33,049 controls) revealed an increased risk of AD in R47H carriers (OR=3.93, 95% CI: 3.15-4.90, P<0.001, I2=0; Fig. S1).
Figure 2
Forest plot for the association of triggering receptor expressed on myeloid cells 2 variant R47H with susceptibility to Alzheimer's disease. OR, odds ratio; M-H, Mantel-Haentzel method; d.f., degree of freedom.
There were three studies reporting on the association between the R47H variant and the susceptibility to early-onset AD (EOAD), which onsets in patients below the age of 65 years. These 3 studies comprised a total of 1,758 cases of EOAD and 2,606 controls (26,27,36). The variant occurred in 2.22% of EOAD cases but only in 0.42% of controls, thus conferring a 4.86-fold increased risk (95% CI: 2.52-9.36, P<0.001, I2=0; Fig. S2) to EOAD.
R62H and risk of AD
The R62H variant was genotyped in 12 datasets comprising 8,525 AD cases and 9,539 controls. Pooled analysis using the fixed-effects model demonstrated a higher risk of AD in carriers than in non-carriers of R62H in the whole sample (OR=1.37, 95% CI: 1.11-1.70, P=0.004, I2=14.2%; Fig. 3) and in Caucasians only (OR=1.39, 95% CI: 1.11-1.74, P=0.004; Fig. S3). However, sensitivity analysis revealed that the association became insignificant (OR=1.17, 95% CI: 0.92-1.50, P=0.192, I2=0; Fig. S4) when the study by Jin et al (11) was excluded, indicating that the study had a significant impact on the pooled effect size and was the main source of heterogeneity. In a subgroup analysis of Caucasian subjects after excluding the study by Jin et al (eight datasets with 13,436 subjects), no association was identified between the R62H variant and AD susceptibility (OR=1.17, 95% CI: 0.90-1.52, P=0.231, I2=0; Fig. S5).
Figure 3
Forest plot for the association of triggering receptor expressed on myeloid cells 2 variant R62H with susceptibility to Alzheimer's disease. OR, odds ratio; M-H, Mantel-Haentzel method; d.f., degree of freedom.
D87N and risk of AD
A total of 11 datasets (10,614 cases and 11,520 controls) were included in the meta-analysis on the association between the D87N variant and AD susceptibility. However, no significant associations were observed in the whole sample (OR=1.61, 95% CI: 0.94-2.76, P=0.081, I2=0; Fig. 4) or in the Caucasian subgroup (OR=1.62, 95% CI: 0.94-2.82, P=0.084, I2=0; Fig. S6).
Figure 4
Forest plot for the association of triggering receptor expressed on myeloid cells 2 variant D87N with susceptibility to Alzheimer's disease. OR, odds ratio; M-H, Mantel-Haentzel method; d.f., degree of freedom.
H157Y and risk of AD
The H157Y variant was rare in AD cases (0.28%) and controls (0.05%). When pooling 8 datasets (10,096 cases and 10,099 controls) together, the analysis indicated that carriers of the H157Y variant were more predisposed to AD (OR=4.22, 95% CI: 1.93-9.21, P<0.001, I2=0; Fig. 5). The variant conferred a 6.20-fold risk to Caucasians (95% CI: 1.60-24.04, P=0.008, I2=0; Fig. S7) from 5 datasets, but was not associated with AD risk in Asians (OR=3.61, 95% CI: 0.61-21.37, P=0.157, I2=55.1%; Fig. S8) from two datasets. Of note, in Caucasians, the variant was only found in ADpatients (13/6,672) but was completely absent in normal controls (0/6,074).
Figure 5
Forest plot for the association of triggering receptor expressed on myeloid cells 2 variant H157Y with susceptibility to Alzheimer's disease. OR, odds ratio; M-H, Mantel-Haentzel method; d.f., degree of freedom.
Sensitivity analysis and publication bias
Sensitivity analysis demonstrated that none of the included datasets significantly affected the pooled effect size except the study of Jin et al (11) for the R62H variant as mentioned above. The funnel plots were symmetric for each mutation (Fig. 6) and Egger's test indicated that there was no publication bias (P>0.05).
Figure 6
Symmetric funnel plots for the meta-analyses of (A) R47H, (B) R62H, (C) D87N and (D) H157Y. OR, odds ratio; M-H, Mantel-Haentzel method; d.f., degree of freedom.
Discussion
The present meta-analysis systematically investigated the genetic associations of four frequently reported rare variants of TREM2 with AD susceptibility. The results indicated that carriers of the R47H, R62H or H157Y variants were more vulnerable to AD.TREM2 encodes a transmembrane immune receptor on microglial cells and exerts its effect by regulating the number of myeloid cells, enhancing phagocytosis and modulating the inflammation response in the CNS (6,41). Its pivotal role in neuroimmunology and neuroinflammation may indicate an implication of TREM2 in the pathogenesis of neurodegenerative disease (41). Association studies have identified rare variants in the coding sequence of TREM2 as susceptibility markers for amyotrophic lateral sclerosis (42), Parkinson's disease and frontotemporal dementia (43), as well as AD. The involvement of TREM2 in the etiology of neurodegenerative diseases, particularly AD, is complex. Numerous studies have revealed that disease-associated variants (mainly R47H) of TREM2 impact amyloid pathology (44), modulate neuritic dystrophy (34), influence tau hyperphosphorylation and aggregation (45) and affect synaptic and neuronal loss (46). Although the pathogenic mechanism of TREM2 in AD may not have been conclusive, TREM2 variants are conclusively susceptibility markers for AD, as demonstrated by the present meta-analysis.R47H is the most frequently investigated variant in TREM2. The substitution of Arg by His leads to a marked reduction in soluble TREM2 levels (47) and the binding ability to cells, APOE and various lipids (45,48,49). However, whether it reduces the mRNA and protein expression in humans remains controversial (50). Furthermore, experiments have revealed that the heterozygous R47H variant may confer a loss of TREM2 function and enhance neuritic dystrophy around plaques, thus increasing the risk of AD (51), which was ascertained by epidemiological investigations in various populations (4,8,40). The R47H variant is enriched in AD cases (1.39%) but less prevalent in cognitively normal controls (0.35%). Most of the identified variants are heterozygous (98%, 514/524) and homozygous variants are detected in AD cases only. The present meta-analysis indicated an approximately 4-fold risk of AD in R47H carriers compared to non-carriers, indicating an effect size similar to that of the APOE Ɛ4 allele (4). However, APOE Ɛ4 is much more prevalent in populations and should still be the main genetic determinant of AD susceptibility.In addition to risk, the R47H variant may also contribute to an earlier age of onset in AD. Slattery et al (32) reported a markedly younger age at onset in patients with the R47H variant than in those without the variant (55.2 vs. 61.7 years on average). Similar results were obtained in Icelandic and Dutch populations (4), implying an association between R47H and susceptibility to EOAD. In the present study, pooled analysis of 3 studies (27,28,37) suggested a higher prevalence (2.22%) of R47H in EOAD cases. Carriers had 4.86-fold risk of developing EOAD, which was even higher than the effect size observed in the total AD samples.The R62H variant disrupts TREM2 recognition of cells and APOE (52) but does not alter the binding to lipid ligands (53). Jin et al (11) determined that it is associated with an increased risk of AD in a population of Americans of European descent (11). The present study revealed a mildly increased risk of AD in individuals harboring the R62H variant (OR=1.38, 95% CI: 1.11-1.70, P=0.004). However, the effect size may be largely attributed to the aforementioned study (11), as indicated by the sensitivity analysis. The association was no longer significant after excluding this study (P=0.192). A significant association was only observed in the study by Jin et al (11), but not in any other study. It should be noted that the study by Jin et al (11) had a large sample size (>3,700 subjects), and thus, it may have more statistical power than the other studies to identify the association of a rare variant with AD susceptibility. Therefore, more studies with a large sample size are required to confirm the association.The D87N variant resides within the ectodomain proximal to the stalk region of TREM2. The mutation leads to enhanced interaction of TREM2 with certain ligands (53). However, the present study did not indicate any association of the variant with AD susceptibility when 11 datasets were pooled together. Of note, D87N was identified in 4 out of 5 patients in a family affected by AD (54). It may be enriched in familial cases in contrast to sporadic cases. Another variant, H157Y, is located on the stalk domain of TREM2 but has no impact on binding activity (53). Specifically, it is located at the TREM2 cleavage site by metalloprotease ADAM metallopeptidase domain 10 and results in increased shedding of TREM2 and reduced cell surface expression of TREM2(55). In addition, in silico programs [SIFT (http://sift.jcvi.org/) and Polyphen2 (http://genetics.bwh.harvard.edu/pph2/)] predicted that the variant may have a deleterious/possibly damaging impact on functions of TREM2. The present analysis, comprising 10,096 cases and 10,099 controls, revealed a significantly increased AD susceptibility in carriers of H157Y, although the variant is rare in AD (0.28%).In the present study, a subgroup analysis was performed with regard to ethnicity, indicating an ethnicity-specific pattern of TREM2 variants. R47H and R62H were mostly identified in Caucasians but rare in Asians. Only one study from Asian (10) reported a low frequency of R47H (0.06%), whereas the other studies did not (18-22). Therefore, R47H and R62H mutations were only associated with an increased risk of AD in Caucasians. Contrary to R47H and R62H, H157Y was more frequent in Asians. Miyashita et al (10) reported 10 carriers out of 4,631 participants, while Jiang et al (12) identified 9 carriers among 2,352 participants. However, the H157Y variant was not associated with AD risk in Japanese (10) but increased the risk in a Han Chinese population (12). Of note, in Caucasians, the H157Y variant was only harbored by AD cases but by none of the controls, indicating that H157Y was more likely to be a causative variant of AD in Caucasians.In summary, the present meta-analysis, involving >73,000 individuals, supported that rare coding variants of TREM2 are associated with AD susceptibility and may be used as predictive markers for neurodegenerative disease.
Authors: Peng Yuan; Carlo Condello; C Dirk Keene; Yaming Wang; Thomas D Bird; Steven M Paul; Wenjie Luo; Marco Colonna; David Baddeley; Jaime Grutzendler Journal: Neuron Date: 2016-10-05 Impact factor: 17.173
Authors: Qingwen Cheng; Jean Danao; Santosh Talreja; Paul Wen; Jun Yin; Ning Sun; Chi-Ming Li; Danny Chui; David Tran; Samir Koirala; Hang Chen; Ian N Foltz; Songli Wang; Shilpa Sambashivan Journal: J Biol Chem Date: 2018-03-29 Impact factor: 5.157
Authors: Margarita Giraldo; Francisco Lopera; Ashley L Siniard; Jason J Corneveaux; Isabelle Schrauwen; Julian Carvajal; Claudia Muñoz; Manuel Ramirez-Restrepo; Chris Gaiteri; Amanda J Myers; Richard J Caselli; Kenneth S Kosik; Eric M Reiman; Matthew J Huentelman Journal: Neurobiol Aging Date: 2013-04-09 Impact factor: 4.673
Authors: Lih-Fen Lue; Christopher T Schmitz; Geidy Serrano; Lucia I Sue; Thomas G Beach; Douglas G Walker Journal: Brain Pathol Date: 2014-11-11 Impact factor: 6.508
Authors: Rita Guerreiro; Aleksandra Wojtas; Jose Bras; Minerva Carrasquillo; Ekaterina Rogaeva; Elisa Majounie; Carlos Cruchaga; Celeste Sassi; John S K Kauwe; Steven Younkin; Lilinaz Hazrati; John Collinge; Jennifer Pocock; Tammaryn Lashley; Julie Williams; Jean-Charles Lambert; Philippe Amouyel; Alison Goate; Rosa Rademakers; Kevin Morgan; John Powell; Peter St George-Hyslop; Andrew Singleton; John Hardy Journal: N Engl J Med Date: 2012-11-14 Impact factor: 91.245
Authors: Mahdi Ghani; Anthony E Lang; Lorne Zinman; Benedetta Nacmias; Sandro Sorbi; Valentina Bessi; Andrea Tedde; Maria Carmela Tartaglia; Ezequiel I Surace; Christine Sato; Danielle Moreno; Zhengrui Xi; Rachel Hung; Mike A Nalls; Andrew Singleton; Peter St George-Hyslop; Ekaterina Rogaeva Journal: Neurobiol Aging Date: 2014-08-01 Impact factor: 4.673
Authors: Boram Kim; EunRan Suh; Aivi T Nguyen; Stefan Prokop; Bailey Mikytuck; Olamide A Olatunji; John L Robinson; Murray Grossman; Jeffrey S Phillips; David J Irwin; Dawn Mechanic-Hamilton; David A Wolk; John Q Trojanowski; Corey T McMillan; Vivianna M Van Deerlin; Edward B Lee Journal: Acta Neuropathol Date: 2022-09-16 Impact factor: 15.887
Authors: Yihan Li; Simon M Laws; Luke A Miles; James S Wiley; Xin Huang; Colin L Masters; Ben J Gu Journal: Cell Mol Life Sci Date: 2021-10-27 Impact factor: 9.207
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6