Suárez-Calvet et al. Molecular Neurodegeneration (2019) 14:1 https://doi.org/10.1186/s13024-018-0301-5 RESEARCH ARTICLE Open Access Early increase of CSF sTREM2 in Alzheimer’s disease is associated with tau related- neurodegeneration but not with amyloid-β pathology Marc Suárez-Calvet1,2,16*, Estrella Morenas-Rodríguez2,3, Gernot Kleinberger1,4, Kai Schlepckow2, Miguel Ángel Araque Caballero5, Nicolai Franzmeier5, Anja Capell1, Katrin Fellerer1, Brigitte Nuscher1, Erden Eren1,6,7, Johannes Levin2,8, Yuetiva Deming9, Laura Piccio10,11, Celeste M. Karch9,11,12, Carlos Cruchaga9,11,12, Leslie M. Shaw13,14, John Q. Trojanowski13,14, Michael Weiner15, Michael Ewers5†, Christian Haass1,2,4*† and for the Alzheimer’s Disease Neuroimaging Initiative Abstract Background: TREM2 is a transmembrane receptor that is predominantly expressed by microglia in the central nervous system. Rare variants in the TREM2 gene increase the risk for late-onset Alzheimer’s disease (AD). Soluble TREM2 (sTREM2) resulting from shedding of the TREM2 ectodomain can be detected in the cerebrospinal fluid (CSF) and is a surrogate measure of TREM2-mediated microglia function. CSF sTREM2 has been previously reported to increase at different clinical stages of AD, however, alterations in relation to Amyloid β-peptide (Aβ) deposition or additional pathological processes in the amyloid cascade (such as tau pathology or neurodegeneration) remain unclear. In the current cross-sectional study, we employed the biomarker-based classification framework recently proposed by the NIA-AA consensus guidelines, in combination with clinical staging, in order to examine the CSF sTREM2 alterations at early asymptomatic and symptomatic stages of AD. Methods: A cross-sectional study of 1027 participants of the Alzheimer’s Disease Imaging Initiative (ADNI) cohort, including 43 subjects carrying TREM2 rare genetic variants, was conducted to measure CSF sTREM2 using a previously validated enzyme-linked immunosorbent assay (ELISA). ADNI participants were classified following the A/T/N framework, which we implemented based on the CSF levels of Aβ1-42 (A), phosphorylated tau (T) and total tau as a marker of neurodegeneration (N), at different clinical stages defined by the clinical dementia rating (CDR) score. Results: CSF sTREM2 differed between TREM2 variants, whereas the p.R47H variant had higher CSF sTREM2, p.L211P had lower CSF sTREM2 than non-carriers. We found that CSF sTREM2 increased in early symptomatic stages of late-onset AD but, unexpectedly, we observed decreased CSF sTREM2 levels at the earliest asymptomatic phase when only abnormal Aβ pathology (A+) but no tau pathology or neurodegeneration (TN-), is present. (Continued on next page) * Correspondence: msuarez@barcelonabeta.org; christian.haass@mail03.med.uni-muenchen.de † Michael Ewers and Christian Haass contributed equally to this work. 1 Chair of Metabolic Biochemistry, Biomedical Center (BMC), Faculty of Medicine, Ludwig-Maximilians-Universität München, Munich, Germany Full list of author information is available at the end of the article © The Author(s). 2019 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Suárez-Calvet et al. Molecular Neurodegeneration (2019) 14:1 Page 2 of 14 (Continued from previous page) Conclusions: Aβ pathology (A) and tau pathology/neurodegeneration (TN) have differing associations with CSF sTREM2. While tau-related neurodegeneration is associated with an increase in CSF sTREM2, Aβ pathology in the absence of downstream tau-related neurodegeneration is associated with a decrease in CSF sTREM2. Keywords: Alzheimer’s disease, Biomarkers, Microglia, Neurodegeneration, Neuroinflammation, Shedding, TREM2 Background early symptomatic stages of late-onset AD [25]. In auto- The triggering receptor expressed on myeloid cells 2 somal dominant AD (ADAD) assessed within the Dom- (TREM2) is an innate immune receptor that is expressed inantly Inherited Alzheimer Network (DIAN) project on the plasma membrane of microglia in the central ner- [26], we demonstrated that CSF sTREM2 was increased vous system (CNS) [1]. TREM2 is involved in key func- in mutation carriers compared to non-carriers five years tions of microglia including phagocytosis, cytokine release, before the estimated years from symptom onset (EYO), lipid sensing and microglia proliferation and migration [2– but with a considerable delay after the development of 6]. TREM2 mutations strongly increase the risk of devel- Aβ pathology, which emerged about 10-15 years earlier. oping Alzheimer’s disease (AD) [7, 8] and other neurode- Together, these studies suggest a complex association of generative diseases including frontotemporal dementia CSF sTREM2 as a function of disease evolution, in (FTD), Parkinson’s disease and amyotrophic lateral scler- which CSF sTREM2 dynamically changes as disease pro- osis [9–12]. Furthermore, homozygous loss-of-function gresses and reaches its highest levels between the later mutations in TREM2 are sufficient to cause Nasu-Hakola asymptomatic and earlier symptomatic stages, when disease (NHD) and FTD-like syndrome [13, 14]. Together, neurodegeneration has already started. this suggests that abnormal TREM2 function plays an es- An important unanswered question in this regard con- sential role across different neurodegenerative diseases. cerns the association between CSF sTREM2 and primary TREM2 is a type-1 transmembrane protein that ma- pathologies including Aβ and tau deposition, as well as tures within the secretory pathway and its ectodomain is neurodegeneration during the course of AD. Therefore, shed at the plasma membrane [2, 15]. Soluble TREM2 we used herein the biomarker-based A/T/N classifica- (sTREM2) accumulates in conditioned media of cultured tion system [27], which is the foundation of the recently cells and in biological fluids such as plasma and cerebro- proposed 2018 NIA-AA research Framework [28]. This spinal fluid (CSF) [2, 16]. Shedding is mediated by classification system consists of three biomarker dimen- ADAM10 and 17 C-terminal to histidine 157 [2, 15, 17– sions including the assessment of Aβ pathology (A), tau 19]. Homozygous mutations causing NHD or FTD-like pathology (T), and neurodegeneration (N). In the syndrome (such as p.T66M) retain misfolded TREM2 in present study, we investigated CSF sTREM2 levels at dif- the endoplasmic reticulum, preventing its maturation ferent AD biomarker-defined groups following the A/T/ and its cleavage on the plasma membrane. Patients bear- N classification and the clinical stage (as defined by the ing these mutations have undetectable levels of sTREM2 clinical dementia rating score, CDR) in participants of in CSF and blood [2, 20, 21]. the well-characterized ADNI study. This approach The fact that TREM2 is selectively expressed in micro- allowed us to test the two main aims of this study. First, glia in the CNS and is associated with AD and neurode- to assess the association of CSF sTREM2 with Aβ path- generation, let us hypothesize that sTREM2 in CSF may ology and its downstream pathological processes (i.e. tau be a marker for microglia function and its response to pathology and neurodegeneration). Second, to assess the Aβ and tau pathology and neurodegeneration. Specific- changes on CSF sTREM2 that occur in the Alzheimer’s ally, sTREM2 may reflect the amount of signaling com- continuum and hence replicate ours and others findings petent TREM2 on the surface of activated microglia. in the ADNI cohort [23–26, 29]. This idea is supported by the fact that the levels of sTrem2 in the brain of an Aβ mouse model correlate Methods with TSPO small animal positron emission tomography ADNI Participants and study design (μPET) signal [22], a marker of microglial activation, and This is a cross-sectional study in which CSF sTREM2 was the fact that a knock-in mouse model bearing the Trem2 measured in 1031 participants of the ADNI project. p.T66M mutation has decreased microglial activity [20]. Among them, 4 individuals did not have the AD core bio- We and others have previously reported changes in markers measurements and were further excluded from the levels of CSF sTREM2 in AD compared to controls the analysis, rendering a study sample of 1027 subjects. [2, 21, 23–26]. Specifically, we found a disease-stage The CSF sTREM2 measurements were uploaded to the dependent increase in CSF sTREM2 peaking within the ADNI database (http://adni.loni.usc.edu) on 16/03/2018 Suárez-Calvet et al. Molecular Neurodegeneration (2019) 14:1 Page 3 of 14 and the data used in this study was downloaded on 21/03/ tau (T) and neurodegeneration (N) biomarkers in the 2018. The ADNI project (http://www.loni.usc.edu/) is a normal range (T- and N-, that is P-tau181P ≤ 21.8 pg/ml multicenter longitudinal study led by Principal Investiga- and T-tau ≤ 245 pg/ml). Participants were classified as tor Michael W Weiner with the main goal to develop and TN positive (TN+) if either aggregated tau (T) or neu- validate biomarkers for subject selection and as surrogate rodegeneration (N) were abnormal (T+ or N+, that is outcome measures in late-onset AD [30]. The institutional P-tau181P > 21.8 pg/ml or T-tau > 245 pg/ml). Only review boards (IRB) of all participating centers approved 5.4% of the individuals of the total differed between the the procedures of the study and all participants or surro- T and N biomarkers groups. gates provided informed consent. Our local IRB (LMU) The combination of the biomarker profile (A/T/N also approved the study. scheme) and the clinical status (CDR) rendered 12 different groups that are displayed in Table 1. We studied CSF sTREM2 in ADNI following two ap- Clinical classification proaches. In a first one, we compared CSF sTREM2 In line with the recently published 2018 NIA-AA “re- between the different A/T/N categories within each search framework” for the diagnosis of Alzheimer’s dis- clinical stage. In a second one, we attempted to ease [28], we assigned each ADNI participant in a model the course of AD with biomarker and group defined by its biomarker profile, as described by clinical-based groups, similar to what was proposed the A/T/N scheme [27], coupled with its cognitive sta- by the 2011 NIA-AA criteria [32–34]. Thus, in this tus, as defined by the CDR score [31]. The A/T/N second approach, we compared the ‘CDR = 0 A-/ scheme comprises 3 biomarker groups: “A” refers to ag- TN-’ group (which corresponds to healthy controls) gregated Aβ, “T” aggregated tau and “N” to neurode- with those biomarker-based groups that fall into the generation. Each biomarker group is binarizied in Alzheimer’s continuum category, that is: ‘Preclinical negative (-) or positive (+) based on whether their bio- AD A+/TN-’, ‘Preclinical AD A+/TN+’, ‘AD CDR = 0.5’ markers are normal or abnormal. In the present study, and ‘AD CDR = 1’. Since our aim was to study the we assigned “A+” to those individuals that had a CSF Alzheimer's continuum, we excluded from this ana- Aβ1-42 < 976.6 pg/ml, “T+” to those individuals with lysis those individuals that fall in the category of sus- P-tau181P > 21.8 pg/ml and “N+” to those individuals pected non-Alzheimer’s pathology (SNAP) [35–38] with T-tau > 245 pg/ml. We merged the aggregated tau and those symptomatic individuals (CDR > 0) that do (T) and neurodegeneration (N) groups in order to de- not have positive biomarkers for both Aβ deposition crease the number of groups to be compared. TN nega- (CSF Aβ1-42) and neurodegeneration/tau pathology tive (TN-) was defined as having both the aggregated (T-tau or P-tau181P). Table 1 Classification of ADNI participants based on the A/T/N framework and clinical stage The ADNI participants were classified based on their clinical stage, as defined by the clinical dementia rating (CDR) score, and the biomarker-based A/T/N framework. The A/T/N framework comprises 3 biomarker groups: A Aβ pathology biomarker status, T tau pathology biomarker status, and N neurodegeneration biomarker status. Each of the biomarkers group have binarized into positive/abnormal (+) or negative/normal according the biomarkers cutoffs. T and N have been merged to simplify the classification and TN- indicates that both T and N are normal and TN+ indicates that T and/or N are abnormal. The numbers shown here are excluding the TREM2 mutation carriers and CSF sTREM2 outliers (as defined as values 3 standard deviations above or below the mean). The colour indicates the different groups used for comparisons in the main text. Healthy controls (n = 122) are depicted in blue, the Alzheimer’s continuum (n = 459) in red and the suspected non-Alzheimer’s pathology (SNAP) group (n = 173) in green. Bold text depicts the groups analysed when modelling the course of AD, namely 'healthy controls', 'Preclinical AD A+/TN -', 'Preclinical AD A+/TN+', 'AD CDR = 0.5' and 'AD CDR = 1' Suárez-Calvet et al. Molecular Neurodegeneration (2019) 14:1 Page 4 of 14 Alzheimer’s disease CSF core biomarkers and CSF sTREM2 antibody against sTREM2 (see Additional file 1: Supple- measurements mentary methods). Second, it was measured by a MSD In the present study, we used the AD CSF core bio- platform-based including an antibody against the HA-tag. markers measurements performed with the Elecsys® This second assay follows the same protocol as the first total-tau CSF, the Elecsys® phosphotau(181P) CSF and one but with the following modifications. The detection the Elecsys® β-amyloid(1–42) CSF immunoassays on a antibody is a monoclonal rat IgG anti-HA peptide se- cobas e 601 instrument [39, 40]. The data is available in quence (YPYDVPDYA), clone 3F10 (Roche, Cat. No. 11 the ‘UPENNBIOMK9.csv’ file in the ADNI database. 867 423 001; 100 ng/mL, 50 μL/well); the secondary anti- These immunoassays are for investigational use only. body is a SULFO-TAG-labeled goat polyclonal anti-rat They are currently under development by Roche Diag- IgG antibody (MSD, Cat. no. R32AH-1; 0.5 μg/mL, 25 μL/ nostics and not commercially available yet. The analyte well). The samples were diluted 1:50 and 1:100 in assay measuring ranges (lower technical limit to upper tech- buffer [0.25% BSA and 0.05% Tween 20 in PBS (pH = nical limit) of these assays are the following: 80 to 1300 7.4)], supplemented with protease inhibitors (Sigma; Cat. pg/ml for total-Tau CSF, 8 to 120 pg/ml for no. P8340) and measured in duplicates for each dilution. phosphor-Tau(181P) CSF, and 200 to 1700 pg/ml for We acquired the electrochemiluminescence response Elecsys® β-Amyloid(1-42) CSF immunoassays. The meas- values using the MESO QuickPlex SQ 120. We compared uring range of the Elecsys® β-Amyloid(1-42) CSF im- the signal of the sTREM2 ELISA with that of the HA-tag munoassay beyond the upper technical limit has not assay for each of the TREM2 rare variants. The percentage been formally established. Therefore use of values above between these two assays renders a relative affinity of the the upper technical limit, which are provided based on sTREM2 ELISA to each of the TREM2 rare variants in re- an extrapolation of the calibration curve, is restricted to lation to its respective HA-tag control. exploratory research purposes and is excluded for clin- ical decision making or for the derivation of medical de- cision points. Statistical analysis CSF sTREM2 measurements were done with a MSD CSF sTREM2 did not follow a normal distribution (Kol- platform-based assay, previously reported and validated mogorov-Smirnov test: P < 0.0001) and were hence log10- [2, 25, 26]. A comprehensive description of the assay is transformed. After transformation, CSF sTREM2 followed shown in Supplementary methods (see Additional file 1: a normal distribution as assessed by Kolmogorov-Smirnov Supplementary methods). The CSF sTREM2 measure- test (P = 0.200) and visual inspection of the histogram. All ments are publicly available in the ADNI database. the statistical analysis described in this study are per- formed with the log10-transformed values. Cell culture and transient transfection of HEK293T cells A one-way analysis of covariance (ANCOVA) was con- HEK293T cells were cultured in DMEM with Glutamax ducted to determine statistically significant differences I supplemented with 10 % (v/v) fetal calf serum (FCS) on CSF sTREM2 between TREM2 rare variants carriers and 1 % penicillin/streptomycin. 24 hours after seeding, and the non-carriers’ individuals adjusting for the effect cells were transiently transfected with equal amounts of of age, followed by a Bonferroni corrected post hoc pair- DNA coding for the different TREM2 variants using Li- wise comparison. Only those groups of TREM2 rare var- pofectamine 2000 as the transfection reagent. TREM2 iants carriers that comprise more than 1 subject were variant constructs were generated by site-directed muta- included in the analysis. genesis (Stratagene) of a TREM2 wt construct bearing The following analyses were conducted excluding out- N- and C-terminal HA and FLAG tags, respectively, as liers’ values of CSF sTREM2, defined as values differing previously described [2]. All constructs were verified by 3 standard deviations from the mean. There were 5 out- DNA sequencing (GATC Biotech). We collected the liers: 2 subjects classified as ‘Preclinical AD A+TN-’ (1 a conditioned media 48 hours after transfection. Cellular TREM2 rare variant carrier and 1 a non-carrier), 1 clas- debris was removed by centrifugation at 4°C (13300 sified as ‘CDR = 0.5 A+TN-’ (TREM2 rare variant car- rpm, 20 min). Supernatants were subsequently frozen at rier), 1 classified as ‘AD CDR = 0.5’ (TREM2 rare variant -20°C until analyses were performed. Cell culture re- carrier), 1 classified as ‘AD CDR = 1’ (non-carrier). In- agents were purchased from Thermo Fisher Scientific cluding or excluding these outliers do not change the unless otherwise noted. findings of this study. We measured the concentrations of HA-labeled TREM2 To study the association of CSF sTREM2 with demo- protein in the HEK293T conditioned media by two differ- graphic and genetic data, we computed a linear regres- ent ELISAs. First, sTREM2 concentrations were deter- sion model with CSF sTREM2 as an outcome variable mined by the same ELISA used to measured sTREM2 in and age, gender and APOE ε4 status as fixed effects. the human CSF samples, which includes a detection Since only age showed to be a significant predictor of Suárez-Calvet et al. Molecular Neurodegeneration (2019) 14:1 Page 5 of 14 CSF sTREM2, the following analyses were conducted in- Results cluding only age as a covariate. Association of CSF sTREM2 with genetic and To test the differences in CSF sTREM2 across bio- demographical data marker profiles in the A/T/N framework, we applied a We studied a total of 1027 participants of the ADNI one-way ANCOVA including age as covariate, followed study. The demographical and clinical characteristics of by Bonferroni corrected post hoc pairwise comparisons. the whole study population are described in Table S1 A similar approach was used to test whether CSF (see Additional file 1). Among the participants studied, sTREM2 changes across the Alzheimer’s continuum. 43 (4.2 %) had a known TREM2 rare variant (see refer- These analyses were performed including or excluding ence [41] for a comprehensive review of the pathogen- individuals carrying a TREM2 rare variant and yielded icity of each variant). The overall mean levels of CSF similar results. sTREM2 of these individuals (M = 3913 pg/ml, SD = Finally, we studied the association between CSF 3548, n = 43) were significantly lower than the rest of sTREM2 and each of the CSF core biomarkers for AD ADNI participants without a TREM2 rare variant (M = (T-tau, P-tau181P, Aβ1-42) with a multiple linear regres- 4136 pg/ml, SD = 2171, n = 984; F1,1024 = 6.77, P = sion adjusted for age. The analysis was conducted separ- 0.009, ηp2 = 0.007; Table 2, Fig. 1) in a one-way ately in the healthy controls, Alzheimer’s continuum and ANCOVA adjusted for age. However, CSF sTREM2 var- SNAP groups. We performed the analysis both including ied considerably between TREM2 variants (F4,1019 = or excluding outliers (defined as AD CSF core bio- 8.79, P < 0.0001, ηp2 = 0.033) and Bonferroni’s post hoc markers 3 standard deviations below or above the group comparisons test indicated that the p.R47H variant [7– mean) in order to exclude that the associations were 12] had significantly higher CSF sTREM2 (P = 0.003) driven by extreme values. The analysis with and without and the p.L211P variant [42, 43] significantly lower CSF outliers rendered similar results. For CSF Aβ1-42, the sTREM2 (P = 0.002) than non-carriers. No differences in analyses were performed using both the truncated values CSF sTREM2 were found between individuals with a at the upper technical limit and the exploratory mea- p.R62H [44, 45] and the p.D87N [7] variants and the surements available based on the extrapolation of the non-carriers. There was a single subject carrying both a calibration curve. In the main text, we report the results p.D87N and p.R62H variants, and another single subject using the extrapolated measurements, but using the carrying a p.H157Y variant, which were not included in truncated ones yielded similar results. the statistical analysis. However, it is worth noting that Statistical analysis was performed in SPSS IBM, ver- the subject carrying a p.H157Y TREM2 rare variant had sion 20.0, and the free statistical software R (http:// relatively high CSF sTREM2 (Table 2, Fig. 1), an obser- www.r-project.org/). Figures were built using GraphPad vation that agrees with our previous findings that the Prism or free statistical software R. All tests were p.H157Y variant, which is located exactly at the cleavage 2-tailed, with a significance level of α = 0.05. site, increases shedding of TREM2 [17]. Given that Table 2 Demographic and clinical characteristics of the individuals carrying a TREM2 rare variant Non-carriers p.R62H p.R47H p.L211P p.D87N p.R62H/D87N p.H157Y (n = 984) (n = 20) (n = 7) (n = 11) (n = 3) (n = 1) (n =1) Age, y 73.1 (7.35) 74.7 (6.47) 73.5 (11.3) 72.8 (4.36) 72.7 (6.17) 66.4 73.1 Female, n (%) 430 (43.7) 11 (55.0) 3 (42.9) 6 (54.5) 0 0 1 APOE ε4 carriers, 467 (47.5) 8 (40.0) 5 (71.4) 2 (18.2) 2 (66.7) 0 0 n (%) Education, y 16.0 (2.78) 15.7 (2.39) 15.6 (2.07) 14.6 (2.54) 17.0 (2.65) 15.0 18.0 CSF biomarkers (pg/ml) T-tau 289 (136) 322 (140) 353 (125) 231 (119) 299 (100) 116 214 P-tau181P 27.9 (14.9) 30.9 (15.6) 36.4 (15.8) 22.2 (12.7) 27.9 (11.5) 9.92 18.1 Aβ1-42 982 (457) 1073 (437) 874 (454) 1246 (515) 944 (670) 925 1700 sTREM2 4136 (2171) 3418 (1786) 8790 (6136) 2386 (1390) 1981 (244) 518 5642 Associated diseases na AD AD, FTD, PD, ALS AD, FTD AD AD AD References na [44, 45] [7–12] [42, 43] [7] [7, 44, 45] [56] Data are expressed as mean and standard deviation (SD) or number (n) and percentage (%), as appropriate. Abbreviations: Aβ1-42 amyloid-β 42, AD Alzheimer’s disease, ALS amyotrophic lateral sclerosis, APOE apolipoprotein E, CSF cerebrospinal fluid, FTD frontotemporal dementia, na non-applicable, PD Parkinson’s disease, P-tau181P tau phosphorylated at threonine 181, T-tau total tau, y years. Suárez-Calvet et al. Molecular Neurodegeneration (2019) 14:1 Page 6 of 14 P = 0.002 P = 0.003 P = 0.002 25000 P = 0.004 P < 0.0001 20000 15000 CSF sTREM2 (pg/ml) 10000 5000 0 non-carriers R62H R47H L211P D87N R62H/ H157Y (n = 984) (n = 20) (n = 7) (n = 11) (n = 3) D87N (n = 1) (n = 1) TREM2 rare variants carriers (n = 43) Fig. 1 CSF sTREM2 in ADNI participants carrying a TREM2 rare variant. Scatter plot representing the levels of CSF sTREM2 in carriers of a TREM2 rare variant, compared to the non-carriers ADNI participants. Solid bars represent the mean and the standard deviation (SD). P-values were assessed by a one-way ANCOVA adjusted for age, followed by Bonferroni corrected post hoc pairwise comparisons between the TREM2 variants carriers’ groups and the non-carriers. We did not include in the comparison those rare variants with only one subject (p.R62H/p.D87N and p.H157Y). TREM2 rare variants may influence CSF sTREM2 (as antibody against the epitope tag. Thus, the decreased described here and in [24]), all the following analyses are CSF sTREM2 found in the p.D87N rare variant should excluding participants carrying these rare variants. be interpreted with caution. Nevertheless, including the TREM2 rare variants carriers In the sample excluding the TREM2 rare variants car- did not change the results. In order to test whether the riers and TREM2 outliers’ values (n = 982, see methods differences in CSF sTREM2 among TREM2 rare variants section), we first assessed which demographic and gen- are influenced by differences in the antibody affinity to etic variables are associated with CSF sTREM2 (de- the mutant sTREM2, we transfected HEK293T cells with scriptives summarized in Table 3). Consistent with an epitope tagged wild type (wt) and mutated TREM2 previous results [21, 23–26], CSF sTREM2 levels were and measured sTREM2 released in the media with the associated with age (β = +0.275, P < 0.0001, ηp2 = same ELISA used for the quantification of CSF sTREM2 0.073), but not with gender (F1,978 = 0.029, P = 0.866, and with an ELISA using an antibody against the epitope ηp2 = 0.00003) or APOE ε4 status (F1,978 = 0.099, P = tag (Additional file 1: Figure S1). This revealed that the 0.753, ηp2 = 0.0001). Consequently, all further analysis p.R47H, p.R62H and p.H157Y TREM2 rare variants included age as a covariate, but not gender or APOE ε4 were detected with a slightly reduced efficiency in our status. ELISA; therefore, the increased levels of CSF sTREM2 found in subjects bearing the p.R47H rare variants are Differences of CSF sTREM2 within the A/T/N classification even slightly underestimated. On the other hand, the of AD p.L211P TREM2 rare variants were detected efficiently, In order to assess the impact of Aβ deposition or the independently of their individual amino acid exchanges. downstream processes of the amyloid cascade (i.e. tau However, the p.D87N TREM2 rare variant was detected pathology and neurodegeneration), we applied the re- with significant less affinity in our ELISA than using the cently proposed A/T/N classification framework of AD, Table 3 Demographic and clinical characteristics of the sample excluding the individuals carrying a TREM2 rare variant CDR = 0 (n = 291) CDR = 0.5 (n = 587) CDR = 1 (n = 104) Suárez-Calvet et al. Molecular Neurodegeneration A-/TN- A+/TN- A+/TN+ A-/TN+ A-/TN- A+/TN- A+/TN+ A-/TN+ A-/TN- (n = 2) A+/TN- (n = 15) A+/TN+ (n = 80) A-/TN+ (n = 7) (n = 122) (n = 52) (n = 45) (n = 72) (n = 118) (n = 93) (n = 282) (n = 94) Age, y 72.5 (5.50) 73.5 (5.95) 76.3 (5.46) 74.6 (6.56) 69.9 (7.56) 72.6 (7.74) 73.3 (7.08) 73.2 (8.16) 89.2 (1.63) 76.3 (6.05) 74.1 (9.28) 79.8 (8.02) Female, n (%) 58 (47.5) 24 (46.2) 24 (53.3) 40 (55.6) 55 (46.6) 21 (22.6) 119 (42.2) 42 (44.7) 0 (0) 5 (33.3) 40 (50.0) 2 (28.6) 20 (38.5) 27 (60.0) 15 (20.8) 24 (20.3) 47 (50.5) 216 (76.6) 32 (34.0) 0 (0) 8 (53.3) 59 (73.8) 1 (14.3) (2019) 14:1 APOE ε4 carriers, n (%) 17 (13.9) Education, y 16.4 (2.79) 16.0 (2.65) 16.5 (2.55) 16.4 (2.57) 16.0 (2.68) 16.1 (2.97) 15.9 (2.90) 15.9 (2.66) 17.0 (0) 16.2 (2.51) 15.1 (2.80) 15.0 (2.08) CSF biomarkers, pg/ml* T-tau 185 (32.2) 167 (40.5) 332 (79.3) 322 (71.9) 183.4 (38.2) 173 (40.0) 381 (136) 338 (111) 234 (6.08) 188 (38.8) 393 (136) 489 (221) P-tau181P 16.2 (2.88) 15.4 (4.07) 33.3 (9.09) 28.9 (7.43) 15.8 (3.32) 15.9 (4.14) 38.9 (14.7) 31.6 (13.7) 19.3 (1.20) 17.4 (3.53) 39.2 (15.0) 43.2 (22.9) Aβ1-42 1456 (223) 723 (196) 717 (168) 1538 (235) 1428 (247) 632 (196) 634 (168) 1427 (290) 1330 (14.9) 538 (185) 570 (159) 1484 (290) sTREM2 3741 (1690) 2835 (1524) 4839 (2240) 5378 (2147) 3436 (1754) 2791 (1292) 4441 (2211) 5478 (2370) 5660 (2008) 3051 (1128) 3967 (2000) 7619 (3539) Data are expressed as mean and standard deviation (SD) or number (n) and percentage (%), as appropriate. *The CSF core biomarkers measurements were performed using the electrochemiluminiscence immunoassays Elecsys Total-tau CSF, phosphor-tau(181P) CSF and β-amyloid(1-42) CSF, which have an upper technical limit of 1300 pg/ml (T-tau), 120pg/ml (P-tau181P) or 1700 pg/ml (Aβ1-42). The values above these limits were truncated to the respective upper technical limit. Abbreviations: A Aβ pathology biomarker status, Aβ1-42 amyloid-β 42, APOE apolipoprotein E, CDR clinical dementia rating, CSF cerebrospinal fluid, N neurodegeneration biomarker status, P-tau181P tau phosphorylated at threonine 181, T tau pathology biomarker status, T-tau total tau, y year Page 7 of 14 Suárez-Calvet et al. Molecular Neurodegeneration (2019) 14:1 Page 8 of 14 which proposes 3 binary biomarker groups [27]: (1) ag- Within the CDR = 0.5 group (i.e. very mild dementia), gregated Aβ (A+/A-), (2) aggregated tau (T+/T-) and (3) there was also a significant difference between the four neurodegeneration (N+/N-). Given that CSF T-tau and biomarker profiles (F3,582 = 40.7, P < 0.0001, ηp2 = CSF P-tau181P were highly correlated, we merged the tau 0.173) and the A+/TN- profile had also the significantly (T) and neurodegeneration (N) groups. ‘TN-’ profile was lowest CSF sTREM2 compared to either other bio- defined as both CSF P-tau181P and T-tau within the nor- marker profile (Fig 2). Both the A+/TN+ and the A-/TN mal range, whereas ‘TN+’ was defined as abnormal + biomarker profiles had significant higher CSF sTREM2 levels of CSF P-tau181P or T-tau. Thus, we compared 4 compared to the A-/TN- profile (Fig. 2). The CDR = 1 different biomarker profiles within each clinical stage, group did not yield a sufficient number of subjects per namely: (1) A-/TN-, (2) A+/TN-, (3) A+/TN+ and (4) A/T/N profile to allow for a group comparison. A-/TN+. We repeated the former analysis also including the Within the CDR = 0 group (i.e. cognitively normal in- individuals with TREM2 rare variants (n = 43; demo- dividuals), a one-way ANCOVA showed a significant graphics in Table S1, see Additional file 1) and this difference between the four biomarker profiles after did not change our conclusions derived from the adjusting for the effect of age (F3,286 = 21.3, P < 0.0001, main analysis (Fig. 2). ηp2 = 0.183). A Bonferroni post hoc test revealed that We also repeated the same analysis classifying the par- the A+/TN- profile had significant lower CSF sTREM2 ticipants based only on their Aβ pathology (A; CSF compared to either other biomarker profile (Fig. 2). Aβ1-42) and tau pathology status (T; CSF P-tau181P), that Only the A-/TN+ profile had significant higher CSF is A/T classification, or based only on their Aβ pathology sTREM2 compared to the normal biomarkers profile (A; CSF Aβ1-42) and neurodegeneration status (N; CSF (i.e. A-/TN-). T-tau), that is A/N classification. The results are shown 20000 P < 0.0001 (P < 0.0001) P < 0.0001 P = 0.0002 (P < 0.0001) (P = 0.0001) P < 0.0001 P < 0.0001 (P < 0.0001) (P < 0.0001) P = 0.0004 P < 0.0001 P = 0.002 P < 0.0001 P < 0.0001 (P = 0.001) (P < 0.0001) (P = 0.001) (P < 0.0001) (P = 0.001) 15000 CSF sTREM2 (pg/ml) 10000 5000 0 A-TN- A+TN- A+TN+ A-TN+ A-TN- A+TN- A+TN+ A-TN+ A-TN- A+TN- A+TN+ A-TN+ (n = 122) (n = 52) (n = 45) (n = 72) (n = 118) (n = 93) (n = 282) (n = 94) (n = 2) (n = 15) (n = 80) (n = 7) CDR = 0 CDR = 0.5 CDR = 1 Fig. 2 CSF sTREM2 in the A/T/N framework. Scatter plot depicting the levels of CSF sTREM2 for each of the four biomarker profiles, as defined by the A/T/N framework, coupled with clinical staging, as defined by CDR. The biomarkers groups T (tau pathology) and N (neurodegeneration) were merged in order to reduce the number of groups to compare. The CDR = 1 stage includes some biomarker profiles will low number of subjects, which precludes performing statistical analysis. They are still shown in the figure for sake of completeness. Each biomarker category is represented in a different colour: healthy controls are depicted in blue, Alzheimer’s continuum category in red, SNAP category in green, and purple depicts biomarker profiles not assigned in a specific category in the present study. The analysis reported in the main text was conducted excluding the TREM2 rare variants carriers, the P-values are reported in bold, and the number of individuals (n) per group indicated. Including these carriers (depicted in yellow) rendered similar results (P-values reported between brackets). Solid bars represent the mean and the standard deviation (SD). P-values were assessed by a one-way ANCOVA adjusted for age, followed by Bonferroni corrected post hoc pairwise comparisons. Abbreviations: A: Aβ pathology biomarker status; AD: Alzheimer’s disease; CDR: clinical dementia rating; CSF: cerebrospinal fluid; N: neurodegeneration biomarker status; SNAP: suspected non-Alzheimer’s pathology; T: tau pathology biomarker status. Suárez-Calvet et al. Molecular Neurodegeneration (2019) 14:1 Page 9 of 14 in Figure S2 (see Additional file 1) and they are similar different between groups after adjusting for the effect to those shown with the A/TN classification of the main of age (F4,575 = 11.5, P < 0.0001, ηp2 = 0.074). A post text. Thus, we conclude that the pathological processes hoc analysis using the Bonferroni criterion for signifi- that are downstream of Aβ pathology, both tau path- cance indicated that the average CSF sTREM2 was sig- ology and neurodegeneration, are associated with in- nificantly higher in the ‘AD CDR = 0.5’ group than in creased CSF sTREM2. the ‘healthy controls’ and ‘Preclinical AD A+TN-’ groups (P = 0.034 and P < 0.0001, respectively; Fig. 3). CSF sTREM2 changes across the Alzheimer’s continuum Similar results were obtained when the individuals car- Next we asked if CSF sTREM2 changes during the rying a TREM2 rare variant were included (Fig. 3). course of the disease as previously described in our Thus, these results replicate our and other groups pre- study on late onset-AD [25]. We modeled the evolution vious findings of increased CSF sTREM2 in early symp- of AD comparing the biomarker-defined groups (Table tomatic stages of late-onset AD in an independent 1) that reflect the temporary course of late-onset AD, sample [23–25, 29]. similar to what was proposed by the previous 2011 NIA-AA diagnostic criteria [32–34]. Thus, we com- CSF sTREM2 is associated with T-tau and P-tau but not pared the ‘healthy controls’ group (highlighted in blue Aβ1-42 in Table 1 and corresponding to the the ‘CDR = 0 A-/ Finally, we studied the associations of CSF sTREM2 TN-’ group), with those belonging to the Alzheimer’s with each of the CSF core biomarkers of AD, that is continuum (highlighted in red in Table 1), which in- T-tau, P-tau181P and Aβ1-42, in linear regression cluded: ‘Preclinical AD A+/TN-’, ‘Preclinical AD A+/TN models adjusted for age. The associations were tested +’, ‘AD CDR = 0.5’ and ‘AD CDR = 1’. A one-way separately in three groups based on their biomarker ANCOVA revealed that CSF sTREM2 was significantly profile (see Table 1): (1) healthy controls, (2) P = 0.034 (P = 0.007) P = 0.001 (P = 0.001) 20000 P < 0.0001 (P < 0.0001) P = 0.002 P < 0.0001 (P = 0.004) (P < 0.0001) 15000 CSF sTREM2 (pg/ml) 10000 5000 0 HC A+/TN- A+/TN+ AD AD (n = 122) (n = 52) (n = 45) CDR = 0.5 CDR = 1 Preclinical AD (n = 282) (n = 80) Alzheimer’s continuum Fig. 3 CSF sTREM2 across the Alzheimer’s continuum. Scatter plot showing the levels of CSF sTREM2 in healthy controls (depicted in blue) and the different stages of the Alzheimer’s continuum (depicted in red). The main analysis was conducted excluding the TREM2 rare variants carriers, the P-values are reported in bold, and the number of individuals (n) per group indicated. Including these carriers (depicted in yellow) rendered similar results (P-values reported between brackets) P-values were assessed by a one-way ANCOVA adjusted for age, followed by Bonferroni corrected post hoc pairwise comparisons. Abbreviations: A: Aβ pathology biomarker status; AD: Alzheimer’s disease; CDR: clinical dementia rating; CSF: cerebrospinal fluid; N: neurodegeneration biomarker status; SNAP: suspected non-Alzheimer’s pathophysiology; T: tau pathology biomarker status. Suárez-Calvet et al. Molecular Neurodegeneration (2019) 14:1 Page 10 of 14 Healthy controls Alzheimer’s continuum SNAP A 15000 β = +0.378 β = +0.418 β = +0.255 P < 0.0001 P < 0.0001 P = 0.0005 (pg/ml) 10000 CSF sTREM2 5000 100 150 200 250 200 400 600 300 400 500 600 CSF T-tau (pg/ml) CSF T-tau (pg/ml) CSF T-tau (pg/ml) B 15000 β = +0.389 β = +0.444 β = +0.210 P < 0.0001 P < 0.0001 P = 0.005 (pg/ml) 10000 CSF sTREM2 5000 10.0 12.5 15.0 17.5 20.0 20 40 60 20 40 60 CSF P-tau (pg/ml) CSF P-tau (pg/ml) CSF P-tau (pg/ml) C 15000 β = +0.125 β = +0.064 β = +0.083 P = 0.152 P = 0.163 P = 0.271 (pg/ml) 10000 CSF sTREM2 5000 1000 1500 2000 200 400 600 800 1000 1000 1500 2000 2500 CSF Aβ1-42 (pg/ml) CSF Aβ1-42 (pg/ml) CSF Aβ1-42 (pg/ml) Fig 4. Association of CSF sTREM2 and AD core CSF biomarkers. Scatter plots representing the associations of CSF sTREM2 with each of the AD CSF core biomarkers: T-tau (a), P-tau181P (b), and Aβ1-42 (c) in three different groups defined by the biomarker profile: healthy controls (blue), Alzheimer’s continuum (red) and SNAP groups (green). The solid lines indicate the regression line and the 95% confidence interval for each of the groups. The standardized regression coefficients (β) and the P-values are shown and were computed using a linear model adjusting for age, and are conducted excluding the outliers values. Including them, did not change the conclusions (see Additional file 1: Table S2). We also performed the analysis including the participants carrying a TREM2 rare variant (depicted in yellow) and the results were similar (see Additional file 1: Table S3). Abbreviations: Aβ1-42: amyloid-beta 42; T-tau: total tau; P-tau: tau phosphorylated at Threonine 181; SNAP: suspected non-Alzheimer’s pathology. individuals of the Alzheimer’s continuum and (3) in Discussion the SNAP groups. Consistent with previous findings, In the present study, we assessed the microglial-activity CSF sTREM2 is associated with T-tau and P-tau181P marker CSF sTREM2 within the early phases of AD. To in the three groups studied (Fig. 4). In contrast, no this end, we applied the biomarker-based A/T/N classifi- significant associations were found between CSF cation in combination with clinical staging [27]. The use sTREM2 and Aβ1-42 (Fig. 4). Including the CSF bio- of this classification system enabled us to unravel the ef- markers outliers (see Additional file 1: Table S2), or fect of Aβ pathology and its downstream processes (i.e. including the individuals carrying a TREM2 rare vari- tau pathology and neurodegeneration) on the levels of ants (see Additional file 1: Table S3), did not change CSF sTREM2. Interestingly, they are differentially associ- our findings. ated with CSF sTREM2. While pure Aβ deposition (as Suárez-Calvet et al. Molecular Neurodegeneration (2019) 14:1 Page 11 of 14 defined here as low CSF Aβ1-42) is associated with de- with evidence of Aβ pathology but without signs of tau creased CSF sTREM2, tau pathology or neurodegenera- pathology or neurodegeneration. We did not observe tion (as defined here as increased CSF P-tau181P or CSF this finding in previous studies, probably due to the low T-tau, respectively) are associated with an increase in number of participants in the preclinical stage of CSF sTREM2. The higher CSF sTREM2 in the SNAP late-onset AD and because we did not apply the A/T/N groups confirm that CSF sTREM2 rises with neurode- classification [25]. Noteworthy, we previously observed generation without Aβ pathology. in ADAD that CSF sTREM2 was lower in ADAD muta- Moreover, we show that the levels of CSF sTREM2 dif- tion carriers than in non-carriers at very early stages fer between TREM2 variants and they are increased in (EYO < -15; CSF T-tau becomes significantly increased the p.R47H TREM2 variant compared to non-carriers, in mutation carriers at EYO = -15), yet statistically but decreased in the p.L211P variant and remained un- non-significant [26]. In contrast, we show here that CSF changed in the p.R62H variant, which is consistent with sTREM2 increases as soon there are signs of we show previous reports [24]. Increased levels of sTREM2 in the here that CSF sTREM2 increases, both in the context of p.R47H variant were somewhat surprising; however, very AD (that is with co-occurrence of Aβ pathology) or in little is known if that variant affects proteolytic process- the SNAP patients (where there is no Aβ underlying ing of TREM2. In contrast, the increased levels of the pathology). Consistent with these findings, CSF sTREM2 p.H157Y variant is in line with our previous findings in is distinctly associated with CSF T-tau and CSF cultured cells demonstrating that this variant increased P-tau181P, but not with CSF Aβ1-42. A word of caution is shedding of TREM2 and therefore decreased TREM2 needed in the SNAP category, given that this is an dependent phagocytosis [17]. A word of caution should heterogenous group that most likely exhibits a non-AD be noted with the p.D87N variant, because the lower related neurodegeneration. Well-powered future studies levels could be due, at least partially, to the lesser affinity should address how CSF sTREM2 changes in neurode- of the antibody to this variant. generative diseases different from AD. Among the strengths of this study are the large and The flexibility of this new classification framework also well-characterized sample size and the use of a reliable enabled us to model the natural history of AD and con- assay to measure sTREM2. Yet, this study has some lim- firm in the ADNI study our previous findings in partici- itations. First, this is a cross-sectional study and the re- pants of several European memory clinics [25] who were sults need to be confirmed in a longitudinal setting. classified using the 2011 NIA-AA criteria [32–34, 36]. Second, we used the CSF biomarkers to classify the ADNI Herein, we demonstrate that, after the initial decrease of participants in the A/T/N classification. Although the role CSF sTREM2 in the ‘Preclinical AD A+/TN-’ group, CSF of CSF biomarkers in AD is well-established and a sTREM2 rise is the ‘Preclinical AD A+/TN+’group and complete A/T/N characterization is possible only with in the early symptomatic stage (CDR = 0.5) of AD, albeit CSF biomarkers, CSF T-tau may not necessarily reflect only statistically significant in the latter group. These neurodegeneration but could result from physiological findings therefore replicate our [25, 26] and other groups production of tau [46]. Here, we found CSF P-tau181P and [23, 24, 29] previous findings in which an increase in CSF T-tau to be highly correlated (only 5.4% of all the CSF sTREM2 in early symptomatic AD was observed. ADNI participants had a discrepant T and N biomarker The mechanism underlying the dynamic changes of group), and consequently we merged the “T” (tau path- CSF sTREM2 throughout the course of the disease still ology) and “N” (neurodegeneration) groups. Importantly, need to be investigated. Several studies have consistently the TN- group had normal levels of both T-tau and demonstrated that microglia upregulate TREM2 expres- P-tau181P, which reasonably ensure that other comorbidi- sion in AD mouse models and in human AD brains [3, ties that may cause neural injury (and hence microglial ac- 47, 48]. Moreover, detailed transcriptomics studies that tivation) are excluded. A plethora of other biofluid and investigated microglia in mouse models of AD and neu- neuroimaging markers (i.e. Aβ and tau PET, blood and rodegeneration showed that TREM2 is upregulated in CSF neurofilament light protein, anatomic MRI and the disease-associated microglia (DAM) [1, 49–53]. This FDG-PET) are in principle applicable for the implementa- is consistent with the finding of increased CSF sTREM2 tion of the A/T/N classification, and future studies using in stages downstream of Aβ accumulation, that is when these biomarkers are needed to confirm our results. tau pathology and/or neurodegeneration occur and micro- The A/T/N classification used herein is a descriptive glia may adopt their disease associated molecular signa- biomarker-based classification that does not assume any ture. We were surprised, however, by the observation of temporal sequence of events in AD and is independent an initial CSF sTREM2 decrease in the only Aβ-pathology of the clinical stage of the disease. By applying this clas- stage, which corresponds to the earliest stage of the dis- sification framework, we found an unexpected observa- ease. The possible mechanisms behind this observation tion, namely a decrease of CSF sTREM2 in individuals are still elusive. However, it has been described that Suárez-Calvet et al. Molecular Neurodegeneration (2019) 14:1 Page 12 of 14 microglia are activated in two steps with an initial be found at: http://adni.loni.usc.edu/wp-content/uploads/how_to_apply/ TREM2-independent process followed by a ADNI_Acknowledgement_List.pdf TREM2-dependent process [49]. The CSF sTREM2 in- Availability data and materials crease observed following the initially low levels may re- The dataset supporting the conclusions of this article is available in the ADNI flect the second step of activation. An alternative site, http://adni.loni.usc.edu/. explanation would be that microglia initially forms a bar- Funding rier around plaques [54, 55] and the sTREM2 released by This work was supported by the Deutsche Forschungsgemeinschaft (DFG) microglia is retained within the plaque, until the barrier within the framework of the Munich Cluster for Systems Neurology (EXC 1010 SyNergy), a DFG funded Koselleck Project (HA1737/16-1 to CH) and the fails, and subsequent neural injury starts. Finally, it could AFTD Biomarker Award (to MSC, JL, ME and CH). MSC received funding from also be argued that subjects with low TREM2-function the European Union’s Horizon 2020 Research and Innovation Program under (and hence lower CSF sTREM2 levels) are more the Marie Sklodowska-Curie action grant agreement No 752310. This work was also supported by grants from the National Institutes of Health prone to experience an accelerated early amyloidogen- (R01AG044546, RF1AG053303, RF1AG058501, and U01AG058922), YD was esis (Parhizkar et al. Nat. Neursci in press) and are supported by a NIMH institutional training grant (T32MH014877). LP was therefore overrepresented in the Preclinical AD A supported by a grant from the Fondazione Italiana Sclerosi Multipla (FISM 2017/R/20). EM was supported by a grant from the Ad-Hoc Committee +TN- group. Nevertheless, we are cognizant of the for Young Neurologist (Spanish Society of Neurology) and Health Institute fact that this is an observational study and the find- Carlos III (funding program for the mobility of the researchers). ings reported herein do not elucidate precise mecha- Authors’ contributions nisms. Further work with longitudinal data is needed MSC, EMR, GK, AC, KF, BN, EE, YD and LP performed the experiments. MSC, to address whether the stage-dependent changes in MAC, NF, YD, ME and CH analyzed and interpreted the data. KS and EMR CSF sTREM2 predict a better or worse clinical performed the cell culture experiments. YD, LP, CMK and CC were involved in extracting the genetic data. JL, LS, JQT, MW contributed with patient outcome. samples and/or data. MSC, ME and CH designed the study and wrote the In conclusion, the present study represents the first manuscript. All authors critically reviewed and approved the final manuscript. attempt to study CSF sTREM2 based on the A/T/N Ethics approval and consent to participate classification framework. We demonstrate in the The study was approved by the Ludwig-Maximilians Universität München ADNI cohort that the increase in CSF sTREM2 which institutional review board (IRB), as well as the IRB of all participating centers occurs in early stages parallels the increase in bio- in ADNI. markers of tau pathology and neurodegeneration. In Consent for publication contrast, Aβ deposition in the absence of tau depos- Not applicable. ition and neurodegeneration is associated with lower Competing interests CSF sTREM2. CH collaborates with DENALI Therapeutics and received speakers honoraria from Novartis and Roche. KS collaborates with DENALI. JL reports to receive consulting fees from Aesku, Axon Neuroscience and Ionis Pharmaceuticals, Additional file speakers’ fees from Bayer Vital and the Willi Gross and non-financial support from Abbvie, outside the submitted work. CC receives research support from Additional file 1: CSF sTREM2 measurement. Figure S1. Effects of rare Biogen, EISAI, Alector and Parabon, and is a member of the advisory board TREM2 variants on antibody affinities. Figure S2. CSF sTREM2 in the A/T of ADx Healthcare. The funders of the study had no role in the collection, and in the A/N classifications. Table S1. Demographic and clinical analysis, or interpretation of data; in the writing of the report; or in the characteristics of the entire sample. Table S2. Associations of CSF decision to submit the paper for publication. The remaining authors declare sTREM2 with AD CSF core biomarkers including the biomarkers that they have no competing of interest. outliers. Table S3. Associations of CSF sTREM2 with AD CSF core biomarkers including subjects carrying a TREM2 rare variant. (DOCX 1992 kb) Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Abbreviations Author details Aβ: Amyloid β-peptide; AD: Alzheimer’s disease;; ADAD: Autosomal-dominant 1 Chair of Metabolic Biochemistry, Biomedical Center (BMC), Faculty of Alzheimer’s disease; CDR: Clinical dementia rating score; CNS: Central Medicine, Ludwig-Maximilians-Universität München, Munich, Germany. nervous system; CSF: Cerebrospinal fluid; EYO: Estimated years from 2 German Center for Neurodegenerative Diseases (DZNE) Munich, Munich, symptom onset; SNAP: Suspected non-Alzheimer’s pathology; Germany. 3Department of Neurology, Institut d’Investigacions Biomèdiques, sTREM2: Soluble triggering receptor expressed on myeloid cells 2; Hospital de la Santa Creu i Sant Pau, Universitat Autònoma de Barcelona, TREM2: Triggering receptor expressed on myeloid cells 2 Barcelona, Catalonia, Spain. 4Munich Cluster for Systems Neurology (SyNergy), Munich, Germany. 5Institute for Stroke and Dementia Research, Acknowledgements Klinikum der Universität München, Ludwig-Maximilians-Universität München, We would like to thank all the researchers and participants in the ADNI Munich, Germany. 6Izmir International Biomedicine and Genome Institute initiative. We also thank Nicole Exner for technical assistance and Nicholas Dokuz Eylul University, Izmir, Turkey. 7Department of Neuroscience, Institute Ashton for critically reading the manuscript and helpful discussion. of Health Sciences, Dokuz Eylul University, Izmir, Turkey. 8Department of Data used in preparation of this article were obtained from the Alzheimer’s Neurology, Ludwig-Maximilians-Universität München, Munich, Germany. 9 Disease Neuroimaging Initiative (ADNI) database (adni.loni.usc.edu). As such, Department of Psychiatry, Washington University School of Medicine, Saint the investigators within the ADNI contributed to the design and Louis, MO, USA. 10Department of Neurology, Washington University School implementation of ADNI and/or provided data but did not participate in of Medicine, St. Louis, MO, USA. 11Hope Center for Neurological Disorders, analysis or writing of this report. A complete listing of ADNI investigators can Washington University in St. Louis, St. Louis, MO, USA. 12Knight Alzheimer’s Suárez-Calvet et al. Molecular Neurodegeneration (2019) 14:1 Page 13 of 14 Disease Research Center, Washington University in St. Louis, St. Louis, MO, 19. Thornton P, Sevalle J, Deery MJ, Fraser G, Zhou Y, Ståhl S, et al. TREM2 USA. 13Department of Pathology and Laboratory Medicine, Perelman School shedding by cleavage at the H157-S158 bond is accelerated for the of Medicine, University of Pennsylvania, Philadelphia, PA, USA. 14Center for Alzheimer’s disease-associated H157Y variant. EMBO Mol Med. 2017;9:1366–78. Neurodegenerative Disease Research, Institute on Aging, Perelman School of 20. Kleinberger G, Brendel M, Mracsko E, Wefers B, Groeneweg L, Xiang X, Medicine, University of Pennsylvania, Philadelphia, PA, USA. 15University of et al. The FTD-like syndrome causing TREM2 T66M mutation impairs California at San Francisco, San Francisco, CA, USA. 16Barcelonaβeta Brain microglia function, brain perfusion, and glucose metabolism. EMBO J. Research Center (BBRC), Pasqual Maragall Foundation, Barcelona, Catalonia, 2017;36:1837–53. Spain. 21. Henjum K, Almdahl IS, Årskog V, Minthon L, Hansson O, Fladby T, et al. Cerebrospinal fluid soluble TREM2 in aging and Alzheimer’s disease. Received: 1 August 2018 Accepted: 4 December 2018 Alzheimers Res Ther. 2016;8:17. 22. Brendel M, Kleinberger G, Probst F, Jaworska A, Overhoff F, Blume T, et al. Increase of TREM2 during Aging of an Alzheimer’s Disease Mouse Model Is Paralleled by Microglial Activation and Amyloidosis. Front Aging Neurosci. References 2017;9:8. 1. Hickman SE, Kingery ND, Ohsumi TK, Borowsky ML, Wang LC, Means TK, 23. Heslegrave A, Heywood W, Paterson R, Magdalinou N, Svensson J, et al. The microglial sensome revealed by direct RNA sequencing. Nat Johansson P, et al. Increased cerebrospinal fluid soluble TREM2 Neurosci. 2013;16:1896–905. concentration in Alzheimer’s disease. Mol Neurodegener. 2016;11:3. 2. Kleinberger G, Yamanishi Y, Suárez-Calvet M, Czirr E, Lohmann E, Cuyvers E, 24. Piccio L, Deming Y, Del-Águila JL, Ghezzi L, Holtzman DM, Fagan AM, et al. et al. TREM2 mutations implicated in neurodegeneration impair cell surface Cerebrospinal fluid soluble TREM2 is higher in Alzheimer disease and transport and phagocytosis. Sci Transl Med. 2014;6:243ra86. associated with mutation status. Acta Neuropathol. 2016;131:925–33. 3. Wang Y, Cella M, Mallinson K, Ulrich JD, Young KL, Robinette ML, et al. 25. Suárez-Calvet M, Kleinberger G, Araque Caballero MÁ, Brendel M, Rominger TREM2 lipid sensing sustains the microglial response in an Alzheimer’s A, Alcolea D, et al. sTREM2 cerebrospinal fluid levels are a potential disease model. Cell. 2015;160:1061–71. biomarker for microglia activity in early-stage Alzheimer’s disease and 4. Colonna M, Wang Y. TREM2 variants: new keys to decipher Alzheimer associate with neuronal injury markers. EMBO Mol Med. 2016;8:466–76. disease pathogenesis. Nat Rev Neurosci. 2016;17:201–7. 26. Suárez-Calvet M, Araque Caballero MÁ, Kleinberger G, Bateman RJ, Fagan 5. Mazaheri F, Snaidero N, Kleinberger G, Madore C, Daria A, Werner G, et al. AM, Morris JC, et al. Early changes in CSF sTREM2 in dominantly inherited TREM2 deficiency impairs chemotaxis and microglial responses to neuronal Alzheimer’s disease occur after amyloid deposition and neuronal injury. Sci injury. EMBO Rep. 2017;18:1186–98. Transl Med. 2016;8:369ra178. 6. Xiang X, Werner G, Bohrmann B, Liesz A, Mazaheri F, Capell A, et al. TREM2 deficiency reduces the efficacy of immunotherapeutic amyloid clearance. 27. Jack CR, Bennett DA, Blennow K, Carrillo MC, Feldman HH, Frisoni GB, et al. EMBO Mol Med. 2016;8:992–1004. A/T/N: An unbiased descriptive classification scheme for Alzheimer disease 7. Guerreiro R, Wojtas A, Bras J, Carrasquillo M, Rogaeva E, Majounie E, et al. biomarkers. Neurology. 2016;87:539–47. TREM2 variants in Alzheimer’s disease. N Engl J Med. 2013;368:117–27. 28. Jack CR, Bennett DA, Blennow K, Carrillo MC, Dunn B, Haeberlein SB, et al. 8. Jonsson T, Stefansson H, Steinberg S, Jonsdottir I, Jonsson PV, Snaedal J, NIA-AA Research Framework: Toward a biological definition of Alzheimer’s et al. Variant of TREM2 Associated with the Risk of Alzheimer’s Disease. N disease. Alzheimer’s Dement. 2018;14:535–62. Engl J Med. 2013;368:107–16. 29. Liu D, Cao B, Zhao Y, Huang H, McIntyre RS, Rosenblat JD, et al. Soluble 9. Borroni B, Ferrari F, Galimberti D, Nacmias B, Barone C, Bagnoli S, et al. TREM2 changes during the clinical course of Alzheimer’s disease: A meta- Heterozygous TREM2 mutations in frontotemporal dementia. Neurobiol analysis. Neurosci Lett. 2018;686:10–6. Aging. 2014;35:934.e7–10 30. Weiner MW, Veitch DP, Aisen PS, Beckett LA, Cairns NJ, Green RC, et al. 10. Cady J, Koval ED, Benitez BA, Zaidman C, Jockel-Balsarotti J, Allred P, et al. Recent publications from the Alzheimer’s Disease Neuroimaging Initiative: TREM2 variant p.R47H as a risk factor for sporadic amyotrophic lateral Reviewing progress toward improved AD clinical trials. Alzheimer’s Dement. sclerosis. JAMA Neurol. 2014;71:449–53. 2017;13:e1–85. 11. Cuyvers E, Bettens K, Philtjens S, Van Langenhove T, Gijselinck I, van der Zee 31. Morris JC. The Clinical Dementia Rating (CDR): current version and scoring J, et al. Investigating the role of rare heterozygous TREM2 variants in rules. Neurology. 1993;43:2412–4. Alzheimer’s disease and frontotemporal dementia. Neurobiol Aging. 2014; 32. Sperling RA, Aisen PS, Beckett LA, Bennett DA, Craft S, Fagan AM, et al. 35:726.e11–9. Toward defining the preclinical stages of Alzheimer’s disease: 12. Rayaprolu S, Mullen B, Baker M, Lynch T, Finger E, Seeley WW, et al. recommendations from the National Institute on Aging-Alzheimer’s TREM2 in neurodegeneration: evidence for association of the p.R47H Association workgroups on diagnostic guidelines for Alzheimer’s disease. variant with frontotemporal dementia and Parkinson’s disease. Mol Alzheimers Dement. 2011;7:280–92. Neurodegener. 2013;8:19. 33. Albert MS, DeKosky ST, Dickson D, Dubois B, Feldman HH, Fox NC, et al. The 13. Guerreiro R, Bilgic B, Guven G, Brás J, Rohrer J, Lohmann E, et al. A novel diagnosis of mild cognitive impairment due to Alzheimer’s disease: compound heterozygous mutation in TREM2 found in a Turkish recommendations from the National Institute on Aging-Alzheimer’s frontotemporal dementia-like family. Neurobiol Aging. 2013;34:2890.e1–5. Association workgroups on diagnostic guidelines for Alzheimer’s disease. 14. Paloneva J, Manninen T, Christman G, Hovanes K, Mandelin J, Adolfsson R, Alzheimers Dement. 2011;7:270–9. et al. Mutations in two genes encoding different subunits of a receptor 34. McKhann GM, Knopman DS, Chertkow H, Hyman BT, Jack CR, Kawas CH, signaling complex result in an identical disease phenotype. Am J Hum et al. The diagnosis of dementia due to Alzheimer’s disease: Genet. 2002;71:656–62. recommendations from the National Institute on Aging-Alzheimer’s 15. Wunderlich P, Glebov K, Kemmerling N, Tien NT, Neumann H, Walter J. Association workgroups on diagnostic guidelines for Alzheimer’s disease. Sequential Proteolytic Processing of the Triggering Receptor Expressed on Alzheimers Dement. 2011;7:263–9. Myeloid Cells-2 (TREM2) Protein by Ectodomain Shedding and γ-Secretase- 35. Caroli A, Prestia A, Galluzzi S, Ferrari C, van der Flier WM, Ossenkoppele R, dependent Intramembranous Cleavage. J Biol Chem. 2013;288:33027–36. et al. Mild cognitive impairment with suspected nonamyloid pathology 16. Piccio L, Buonsanti C, Cella M, Tassi I, Schmidt RE, Fenoglio C, et al. (SNAP): Prediction of progression. Neurology. 2015;84:508–15. Identification of soluble TREM-2 in the cerebrospinal fluid and its association 36. Jack CR, Knopman DS, Weigand SD, Wiste HJ, Vemuri P, Lowe V, et al. An with multiple sclerosis and CNS inflammation. Brain. 2008;131:3081–91. operational approach to National Institute on Aging-Alzheimer’s Association 17. Schlepckow K, Kleinberger G, Fukumori A, Feederle R, Lichtenthaler SF, criteria for preclinical Alzheimer disease. Ann Neurol. 2012;71:765–75. Steiner H, et al. An Alzheimer-associated TREM2 variant occurs at the ADAM 37. Dani M, Brooks DJ, Edison P. Suspected non-Alzheimer’s pathology – Is it cleavage site and affects shedding and phagocytic function. EMBO Mol non-Alzheimer’s or non-amyloid? Ageing Res Rev. 2017;36:20–31. Med. 2017;9:1356–65. 38. Jack CR, Knopman DS, Chételat G, Dickson D, Fagan AM, Frisoni GB, et al. 18. Feuerbach D, Schindler P, Barske C, Joller S, Beng-Louka E, Worringer KA, Suspected non-Alzheimer disease pathophysiology — concept and et al. ADAM17 is the main sheddase for the generation of human triggering controversy. Nat Rev Neurol. 2016;12:117–24. receptor expressed in myeloid cells (hTREM2) ectodomain and cleaves 39. Hansson O, Seibyl J, Stomrud E, Zetterberg H, Trojanowski JQ, Bittner T, TREM2 after Histidine 157. Neurosci Lett. 2017;660:109–14. et al. CSF biomarkers of Alzheimer’s disease concord with amyloid-β PET Suárez-Calvet et al. Molecular Neurodegeneration (2019) 14:1 Page 14 of 14 and predict clinical progression: A study of fully automated immunoassays in BioFINDER and ADNI cohorts. Alzheimer’s Dement. 2018;14:1470–81. 40. Bittner T, Zetterberg H, Teunissen CE, Ostlund RE, Militello M, Andreasson U, et al. Technical performance of a novel, fully automated electrochemiluminescence immunoassay for the quantitation of β-amyloid (1–42) in human cerebrospinal fluid. Alzheimer’s Dement. 2016;12:517–26. 41. Jay TR, von Saucken VE, Landreth GE. TREM2 in Neurodegenerative Diseases. Mol Neurodegener. 2017;12:56. 42. Jin SC, Carrasquillo MM, Benitez BA, Skorupa T, Carrell D, Patel D, et al. TREM2 is associated with increased risk for Alzheimer’s disease in African Americans. Mol Neurodegener. 2015;10:19. 43. Thelen M, Razquin C, Hernández I, Gorostidi A, Sánchez-Valle R, Ortega- Cubero S, et al. Investigation of the role of rare TREM2 variants in frontotemporal dementia subtypes. Neurobiol Aging. 2014;35:2657.e13–19. 44. Roussos P, Katsel P, Fam P, Tan W, Purohit DP, Haroutunian V. The triggering receptor expressed on myeloid cells 2 (TREM2) is associated with enhanced inflammation, neuropathological lesions and increased risk for Alzheimer’s dementia. Alzheimer’s Dement. 2015;11:1163–70. 45. Lu Y, Liu W, Wang X. TREM2 variants and risk of Alzheimer’s disease: a meta- analysis. Neurol Sci. 2015;36:1881–8. 46. Sato C, Barthélemy NR, Mawuenyega KG, Patterson BW, Gordon BA, Jockel- Balsarotti J, et al. Tau Kinetics in Neurons and the Human Central Nervous System. Neuron. 2018;97:1284–98 e7. 47. Frank S, Burbach GJ, Bonin M, Walter M, Streit W, Bechmann I, et al. TREM2 is upregulated in amyloid plaque-associated microglia in aged APP23 transgenic mice. Glia. 2008;56:1438–47. 48. Lue LF, Schmitz CT, Serrano G, Sue LI, Beach TG, Walker DG. TREM2 Protein Expression Changes Correlate with Alzheimer’s Disease Neurodegenerative Pathologies in Post-Mortem Temporal Cortices. Brain Pathol. 2015;25:469–80. 49. Keren-Shaul H, Spinrad A, Weiner A, Matcovitch-Natan O, Dvir-Szternfeld R, Ulland TK, et al. A Unique Microglia Type Associated with Restricting Development of Alzheimer’s Disease. Cell. 2017;169:1276–90 e17. 50. Krasemann S, Madore C, Cialic R, Baufeld C, Calcagno N, El Fatimy R, et al. The TREM2-APOE Pathway Drives the Transcriptional Phenotype of Dysfunctional Microglia in Neurodegenerative Diseases. Immunity. 2017;47: 566–81 e9. 51. Holtman IR, Raj DD, Miller JA, Schaafsma W, Yin Z, Brouwer N, et al. Induction of a common microglia gene expression signature by aging and neurodegenerative conditions: a co-expression meta-analysis. Acta Neuropathol Commun. 2015;3:31. 52. Chiu IM, Morimoto ETA, Goodarzi H, Liao JT, O’Keeffe S, Phatnani HP, et al. A neurodegeneration-specific gene-expression signature of acutely isolated microglia from an amyotrophic lateral sclerosis mouse model. Cell Rep. 2013;4:385–401. 53. Orre M, Kamphuis W, Osborn LM, Jansen AHP, Kooijman L, Bossers K, et al. Isolation of glia from Alzheimer’s mice reveals inflammation and dysfunction. Neurobiol Aging. 2014;35:2746–60. 54. Condello C, Yuan P, Schain A, Grutzendler J. Microglia constitute a barrier that prevents neurotoxic protofibrillar Aβ42 hotspots around plaques. Nat Commun. 2015;6:6176. 55. Wang Y, Ulland TK, Ulrich JD, Song W, Tzaferis JA, Hole JT, et al. TREM2- mediated early microglial response limits diffusion and toxicity of amyloid plaques. J Exp Med. 2016;213:667–75. 56. Jiang T, Tan L, Chen Q, Tan MS, Zhou JS, Zhu XC, et al. A rare coding variant in TREM2 increases risk for Alzheimer’s disease in Han Chinese. Neurobiol Aging. 2016;42:217.e1–3.