TREM2 mRNA levels are upregulated in Alzheimer’s disease hippocampus

We first measured TREM2 mRNA levels in hippocampal samples from Alzheimer’s disease (AD) cases and controls by RT-qPCR. Four samples did not pass the RNA quality threshold so were not included in the experiments (see TREM2 mRNA expression analysis in the “Methods” section). Eventually, 26 AD cases were compared to 12 controls. None of the subjects included in the study was carrying the TREM2 rs75932628-T variant, in accordance with the low frequency of the variant allele in the European ancestry population [7]. A 3.4-fold increase in TREM2 mRNA levels was observed in the hippocampus of AD cases compared to controls (mean ± SD mRNA levels in AD versus controls: 6.65 ± 4.30 % versus 1.73 ± 1.24 %; p = 1.1E-05) (Fig. 1a). Next, a disease-staging analysis was conducted to investigate changes of TREM2 mRNA levels considering AD severity. We found that TREM2 mRNA levels significantly increased across AD stages (p = 0.003; Fig. 1b). Games-Howell post hoc analysis showed that TREM2 mRNA expression was significantly different between control and Braak stages III–IV (p = 0.003) and between control and Braak stages V–VI (p = 0.035) (Fig. 1b).

We further tested if differences in TREM2 mRNA levels resulted from an increase in a particular type of TREM2 splice variant. Two different types of TREM2 transcripts were identified in the Reference Sequence (RefSeq) database, i.e., TREM2 transcript variant 1 (NM_018965) and TREM2 transcript variant 2 (NM_001271821) (Additional file 1: Figure S1.A). Next, two sets of primers that separately amplify each TREM2 splice variant were designed (Additional file 1: Table S1 and Figure S1.A). Interestingly, we found that mRNA levels of each of both transcripts were significantly higher in the hippocampus from AD patients compared to controls (Additional file 1: Figure S1.B). Nonetheless, the most prominent splice variant in the hippocampal tissue was TREM2 transcript variant 1 (NM_018965), suggesting that the majority of mRNA changes may be accounted for by an increase in that particular splice variant.

TREM2 mRNA levels positively correlate with β-amyloid and p-tau burden

Next, TREM2 mRNA expression was examined in more detail according to the neuropathological AD-related changes measured and recorded from the hippocampus sections. The amyloid plaque score (APS) and average area of both β-amyloid deposition and hyperphosphorylated tau (p-tau) were quantitatively measured as described in the “Methods” section by using ImageJ software [23] (Additional file 1: Figure S2) in the whole set of 30 AD cases and 12 controls included in the study. Interestingly, a significant association between APS and TREM2 mRNA levels was found in the human hippocampus (r = 0.593, p = 1.4E-04). In addition, the average area of β-amyloid burden in the hippocampus was also correlated with TREM2 mRNA levels (r = 0.602, p = 1.0E-04), suggesting that TREM2 mRNA expression parallels the molecular changes induced by β-amyloid deposition. Since we had found an increased TREM2 mRNA expression across Braak stages, a positive correlation with p-tau deposits was expected. Indeed, the average area of p-tau deposition, which included neurofibrillary tangles, neuropil threads, and neuritic plaques, was positively correlated to TREM2 mRNA levels (r = 0.509, p = 0.002).

DNA methylation in the TREM2 transciption start site (TSS)-associated region is increased in AD cases compared to controls

To further explore if differences in TREM2 mRNA levels between AD cases and controls were related to DNA methylation changes, bisulfite cloning sequencing was conducted in a subset of six controls and 10 Alzheimer’s hippocampal samples. Bisulfite cloning sequencing assesses changes in DNA methylation which includes modifications in 5-methylcytosine (5meC) and 5-hydroxymethylcytosine (5hmC). There were no significant differences between control and AD groups regarding age (52 ± 27.9 years versus 74.8 ± 13.7; p value = 0.107) or gender (male percentage) (62.5 versus 62.5 %; p value = 1.0). Moreover, PMI was similar in both groups (7.1 ± 2.9 h versus 7.6 ± 5.6 h; p value = 0.821).

Percentage of DNA methylation at seven individual cytosine guanine dinucleotides (CpGs) in the TSS-associated region of TREM2 gene (Fig. 2a) was measured by sequencing a minimum of 20 clones per individual. Then, average DNA methylation for the entire amplicon was calculated in each individual. Following that approach, a significant increase in DNA methylation at the TSS-associated region of TREM2 gene was observed in the AD cases group compared to the control group (76.2 ± 15.5 % versus 57.9± 17.1 %; p = 0.0016) (Fig. 2b). All seven interrogated CpGs at the TSS-associated region showed a significant (p < 0.05) gain in DNA methylation in AD cases compared to controls (Table 2). After Bonferroni correction, differences between AD and controls in the average DNA methylation for the whole amplicon and DNA methylation in the first CpG of the amplicon (CpG at position 24) remained significant (adjusted p value = 0.00625; Table 2). Interestingly, this result did not follow the classical model of epigenetic regulation where gain in methylation in the promoter region correlates with a decrease in gene expression. To improve the specificity of our assay, we used a negative control by bisulfite sequencing an amplicon located at −874 bp relative to TREM2 transcription start site (Additional file 1: Figure S3). There were no statistically significant differences between AD and control group (90.7 ± 3.9 versus 92.5 ± 6.45; p value = 0.664) for this particular region. We also wanted to test the relationship between average DNA methylation and mRNA expression for the TREM2 gene. No correlation was found between DNA methylation percentage measured by bisulfite sequencing and TREM2 mRNA levels in our set of samples (r = 0.391, p value = 0.134).

CpG position	24	65	129	175	236	269	471	Total	SD
MR control group. %	60.4	68.10	47.9	50.00	62.5	31.2	85.4	57.91	17.1
MR AD group. %	83.7	85.4	66.7	70.7	80.5	49.2	96.7	76.16	15.5
Gain in methylation. %	23.3	17.3	18.8	20.7	18.00	18.00	11.3	18.25	-
p value	0.0015*	0.0163	0.0353	0.0130	0.0180	0.0400	0.0121	0.0016*	-

CpG position denotes the position of the dinucleotide CpG within the amplicon. Total denotes the averaged methylation percentage for the whole amplicon. MR methylation ratio, AD Alzheimer’s disease, SD standard deviation. *significant after Bonferroni correction (adjusted p value = 0.00625)

5hmC enrichment in the TREM2 gene body correlates with TREM2 mRNA levels in the AD hippocampus

Next, we asked whether DNA methylation differences observed in AD cases were indeed the result of 5hmC changes, as higher 5hmC levels at specific sites have been related to increased gene expression [24, 25]. Since bisulfite cloning sequencing is unable to distinguish between 5mC and 5hmC, we performed 5-hydroymethylated DNA immunoprecipitation (5hMeDIP) to selectively enrich DNA fragments containing 5hmC, followed by quantification by RT-qPCR, in a group of 12 AD cases and 5 controls. Following that procedure, we were able to estimate 5hmC levels not only at the TSS-associated region but also at exon 2 and at the 3′ end of the gene. Noticeably, amplicon covering exon 2 included the rs75932628-T variant associated with AD in GWAS (Fig. 3a). No statistically significant differences in 5hmC enrichment were found between groups at TREM2 TSS-associated region (IP/INPUT fold change enrichment ± SD in AD versus controls: 3.52 ± 5.48 versus 1.69 ± 0.95; p value = 0.130), exon 2 (AD versus controls: 18.70 ± 12.73 versus 16.73 ± 5.14; p value = 0.383) and 3′end (AD versus controls: 9.76 ± 7.72 versus 5.77 ± 2.91; p value = 0.633) (Fig. 3b).

Interestingly, a significant positive correlation was found between 5hmC enrichment in exon 2 and TREM2 mRNA levels (r = 0.669; p = 0.005) (Fig. 3c) for the whole set of samples. When considering only the AD hippocampal samples, correlation between 5hmC enrichment in exon 2 and TREM2 mRNA expression was found to be even stronger (r = 0.771; p = 0.005) (Fig. 3d).

Human brain samples and neuropathological examination

We conducted an observational, case-control study comparing postmortem hippocampal samples from AD patients and control hippocampus. Frozen postmortem hippocampal samples from 30 Alzheimer’s disease (AD) cases and 12 controls were provided by the Navarrabiomed Brain Bank. After death, half brain specimens from donors were cryopreserved at −80 °C.

Neuropathological examination was completed following the usual recommendations [40]. Assessment of β-amyloid deposit was carried out by immunohistochemical staining of paraffin-embedded sections (3- to 5-μm thick) with a mouse monoclonal (S6F/3D) anti β-amyloid antibody (Leica Biosystems Newcastle Ltd, Newcastle upon Tyne, UK). Evaluation of neurofibrillary pathology was performed with a mouse monoclonal antibody anti-human PHF-TAU, clone AT-8, (Tau AT8) (Innogenetics, Gent, Belgium), which identifies hyperphosphorylated tau (p-tau) [41]. The reaction product was visualized using an automated slide immunostainer (Leica Bond Max) with Bond Polymer Refine Detection (Leica Biosystems Newcastle Ltd).

Subjects received diagnosis of pathophysiologically proved AD dementia according to the updated National Institute on Aging-Alzheimer’s Association guidelines [42]. Subjects were further classified, based on neurofibrillary pathology, as follows: control (n = 12); Braak stages I–II (n = 3); Braak stages III–IV (n = 17); and Braak stages V–VI (n = 10) [41, 43]. Importantly, to avoid spurious associations, those individuals showing coincident protein deposits different from p-tau or β-amyloid were not eligible for the study. This approach maximizes chances of finding true associations with AD, even though reducing the final sample size. Neuropathological and demographic features of subjects, including age, gender, and postmortem interval are listed in Additional file 1: Table S2. There were significant age differences between the control and AD group (50.7 ± 21.5 years versus 82.3 ± 11.3 years, p < 0.01), and a statistical trend was found towards higher proportion of male in the control group (66.7versus 34.5 %, p = 0.87). The postmortem interval (PMI) was not significantly different between groups (8.2 ± 4.2 h in the control group versus 7.9 ± 7.1 h in the AD group; p = 0.91).

Ethics, consent, and permissions

The Ethics Committee of the “Complejo Hospitalario de Navarra” approved the use of human subjects for this study (Pyto 90/2014). Written informed consent was obtained from all subjects or next of kin.

Genotyping of TREM2 variant rs75932628-T

We wanted to assess whether any subject included in our study was harboring the TREM2 rare variant rs75932628-T. To that end, we genotyped the TREM2 variant rs75932628-T as it has been previously described [44]. In brief, a PCR reaction was carried out to amplify a 172 base pair (bp) amplicon spanning the rs75932628 variant. Next, a restriction fragment length polymorphism analysis was performed by using the restriction enzyme HhaI (Thermo Fisher Scientific Inc., Waltham, MA, USA). The change from C to T in rs75932628 results in loss of the HhaI restriction enzyme site. When separated by electrophoresis, digested PCR fragments from homozygous wild-type subjects show two bands of 89- and 84-bp length, whereas heterozygous subjects harboring TREM2 variant rs75932628-T display an additional 172 bp band, corresponding to the allele that has lost the restriction enzyme site.

TREM2 mRNA expression analysis

Total RNA was isolated from hippocampus homogenates using RNeasy Lipid Tissue Mini kit (QIAGEN, Redwood City, CA, USA), following manufacturer’s instructions. Genomic DNA was removed with a recombinant DNase (TURBO DNA-free™ Kit, Ambion, Inc., Austin, TX, USA). RNA integrity was checked by 1.25 % agarose gel electrophoresis under denaturing conditions. Concentration and purity of RNA were both evaluated with NanoDrop spectrophotometer. Only those RNA samples showing 260 nm/280 nm absorbance ratios between 1.8 and 2.2 and 260 nm/230 nm absorbance ratios higher than 1.8 were considered high enough quality to be included in the study. Complementary DNA (cDNA) was reverse transcribed from 1500 ng total RNA per sample with SuperScript® III First-Strand Synthesis Reverse Transcriptase (Invitrogen, Carlsbad, CA, USA) and primed with a mixture of oligo-d (T) and random primers.

Real-time qPCR (RT-qPCR) reactions were performed in triplicate for each sample using Power SYBR® Green PCR Master Mix (Invitrogen, Carlsbad, CA, USA) in a 7300 real-time PCR System (Applied Biosystems, Foster City, CA, USA). Experiments were repeated twice within independent cDNA sets. Sequences of primer pair were designed by using real-time PCR tool (IDT, Coralville, IA, USA) and are listed in Additional file 1: Table S1. The PCR amplicon sizes range from 116 to 184 bp. All the primers used in RT-qPCR assays were tested for specificity. During designing the primer sets, a Basic Local Alignment Search Tool (BLAST) versus entire mRNA RefSeq database was performed to verify the correct and unique alignment of the fragment. Moreover, dissociation curves were obtained for each assay to check that curves show a single peak with no shoulder. After amplification, we also checked that RT-qPCR reaction had generated a single amplicon of the correct size by performing agarose gel electrophoresis.

The thermal cycling conditions consisted of an initial denaturation step at 95 °C for 10 min followed by 40 cycles of 15 s at 95 °C and 1 min at 60 °C. The geometric mean of ACTB and HPRT housekeeping genes was used to normalize TREM2 mRNA levels [45], and non-template reactions were included as negative controls in each run. Relative expression level of TREM2 mRNA in a particular sample was calculated by the delta delta-CT method, as previously described [46].

Quantitative assessment of β-amyloid and p-tau deposits in brain tissues

In order to quantitatively assess the β-amyloid and p-tau burden for further statistical analysis, we applied a method to quantify protein deposits. This method generates a numeric measurement that reflects the extent of β-amyloid and p-tau deposition. In the case of p-tau, not only neurofibrillary tangles but also neuritic plaques and neuropil threads are taken into account (Additional file 1: Figure S2). Sections of the hippocampus were examined after performing immunostaining with anti β-amyloid and anti-p-tau antibodies as described above in Human brain samples and neuropathological examination. Three pictures were obtained for each immunostained section by using an Olympus BX51 microscope at ×10 magnification power. Focal deposit of β-amyloid, as described by Braak & Braak (neuritic, immature, and compact plaque) [43], was manually determined and was further edited and analyzed with the ImageJ software (figure e-1). Then, β-amyloid plaque count, referred to as amyloid plaque score (APS) (Additional file 1: Figure S2) and total area of β-amyloid deposition were automatically measured by ImageJ and averaged for each section. Regarding p-tau deposit, representative pictures were analyzed with ImageJ software in order to obtain an average quantitative measure of the global p-tau deposit for each section (Additional file 1: Figure S2).

Methylation in the TSS-associated region of TREM2 by bisulfite cloning sequencing

Genomic DNA was isolated from hippocampal tissue by standard methods [47]. Next, 500 ng of genomic DNA was bisulfite converted using the EpiTect Bisulfite Kit (QIAGEN, Redwood City, CA, USA) according to the manufacturer’s instructions. An amplicon spanning the TSS-associated region of TREM2 was amplified by PCR (Fig. 2a). Primer pair sequences were designed by MethPrimer [48] and are listed in Additional file 1: Table S1. Genomic map of the amplicon was drawn using the UCSC Genome Browser [49]. Next, PCR products were cloned using the TopoTA Cloning System (Invitrogen, Carlsbad, CA, USA), and a minimum of 20 independent clones were sequenced for each examined subject by Sanger sequencing [50].

Determination of 5hmC in hippocampus samples

We evaluate 5-hydroxymethycytosine (5hmC) by performing 5-hydroxymethylated DNA immunoprecipitation combined with RT-qPCR (5hMeDIP-RT-qPCR). Genomic DNA (1500 ng) was sonicated to obtain fragments of 150–500 base pair length that were assessed by 1 % agarose gel electrophoretic analysis. Next, DNA containing 5hmC was enriched using the EpiQuik Hydroxymethylated DNA Immunoprecipitation (hMeDIP) Kit (Epigentek, Farmingdale, NY, USA), according to manufacturer’s recommendations. Primer pair sequences are listed in Additional file 1: Table S1. Serial dilutions of samples were performed to determine amplification efficiency for each primer pair.

Real-time qPCR (RT-qPCR) reactions were performed in triplicate for each sample using Power SYBR® Green PCR Master Mix (Invitrogen, Carlsbad, CA, USA) in a 7300 real-time PCR System (Applied Biosystems, Foster City, CA, USA). The PCR amplicon sizes range from 97 to 108 bp. All the primers used in RT-qPCR assays were tested for specificity as indicated above. The amplification reaction program included an initial step of 95 °C for 10 min, followed by 40 cycles of 95 °C for 15 s, 60 °C for 1 min, and a final melt curve analysis step. DNA control from the kit was included in the experiments. Fold enrichment was calculated as the ratio of amplification efficiency of each immunoprecipitated sample over the efficiency of non-immune IgG as of the following the formula: 2^dCt, where dCt = Ct^INPUT – Ct^hMeDIP, and further normalized to the lowest detectable sample.

Statistical data analysis

Statistical analysis was performed with SPSS software version 21.0 (IBM, Inc., USA). Before performing differential analysis, we checked that all continuous variables showed a normal distribution, as per one-sample Kolgomorov-Smirnov test and the normal quantil-quantil (QQ) plots. Summarized data from continuous variables were expressed as mean ± standard deviation. Significance level for all comparisons was set at p value <0.05. Statistical significance for intergroup differences was assessed by the Pearson’s chi-square test for categorical variables and the independent sample t test for continuous variables. After finding that variances were unequal by Levene’s test, one-way analysis of variance (ANOVA) followed by Games-Howell post hoc analysis was used to analyze differences in the expression levels of TREM2 mRNA between stage groups. A logistic regression model (ENTER method) was fit to assess the independent association of TREM2 mRNA levels with AD status, using gender and age as covariates. The Pearson product-moment correlation coefficient analysis was used to correlate TREM2 mRNA expression levels with neuropathological changes. Methylation ratio was calculated as the ratio of methylated CpGs over the total number of CpGs assessed for each CpG site. Gain in methylation was calculated as the subtraction of percentage of methylation between the AD group and control group. Although methylation is not completely independent for each CpG site within the same amplicon, a conservative approach to avoid false positive results were used by applying Bonferroni correction. Difference between two bisulfite sequence groups was evaluated with Mann-Whitney U test. GraphPad Prism version 6.00 for Windows (GraphPad Software, La Jolla, CA, USA) was used to draw the graphs except for methylation figures that were drawn by QUMA software [51].