Alzheimer’s & Dementia - (2016) 1-7 55 1 56 2 57 3 58 4 59 5 Alzheimer’s disease-associated TREM2 variants exhibit either decreased 60 6 61 7 or increased ligand-dependent activation 62 8 63 64 9 a b b c a 10 Q5 Wilbur Song , Basavaraj Hooli , Kristina Mullin , Sheng Chih Jin , Marina Cella , 65 11 Tyler K. Ullanda, Yaming Wanga,1, Rudolph Tanzib,**, Marco Colonnaa,* 66 12 67 a 13 Department of Pathology and Immunology, Washington University School of Medicine, St. Louis, MO, USA 68 b 14 Department of Neurology, Harvard Medical School, and Genetics and Aging Research Unit, Massachusetts General Hospital, Charlestown, MA, USA 69 c 15 Department of Genetics, Yale University School of Medicine, New Haven, CT, USA 70 16 71 17 72 18 Abstract Introduction: TREM2 is a lipid-sensing activating receptor on microglia known to be important for 73 19 Alzheimer’s disease (AD), but whether it plays a beneficial or detrimental role in disease pathogen- 74 75 20 esis is controversial. 76 21 Methods: We analyzed AD risk of TREM2 variants in the NIMH AD Genetics Initiative Study and 22 77 AD Sequencing Project. We compared each variant’s risk and functional impact by a reporter assay. 23 78 24 Finally, we analyzed expression of TREM2 on human monocytes. 79 25 Results: We provide more evidence for increased AD risk associated with several TREM2 variants, 80 26 and show that these variants decreased or markedly increased binding to TREM2 ligands. We identify 81 27 HDL and LDL as novel TREM2 ligands. We also show that TREM2 expression in human monocytes 82 28 is minimal compared to monocyte-derived dendritic cells. 83 29 Discussion: Our results suggest that TREM2 signaling helps protect against AD but can cause harm 84 30 in excess, supporting the idea that proper TREM2 function is important to counteract disease progres- 85 86 31 sion. 87 32 Ó 2016 Published by Elsevier Inc. on behalf of the Alzheimer’s Association. 33 88 34 89 90 35 Keywords: Alzheimer’s disease; Microglia; TREM2; Monocyte; HDL; LDL; Lipoprotein 36 91 37 92 38 93 39 94 40 1. Introduction disease and switching from protective to harmful. Recently, 95 41 genetic studies have identified several AD-associated poly- 96 The role of microglia in the pathogenesis of Alzheimer’s 42 morphisms in microglia-expressed such as TREM2 97 disease (AD) has been a subject of intense study for many 98 43 [2,3], CD33 [4], and CR1 [5]. Among these loci, the greatest 44 years. In response to b-amyloid (Ab) accumulation, micro- 99 risk is associated with the rs75932628T polymorphism in 45 glia aggregate and participate not only in Ab and tau clear- 100 TREM2 encoding the R47H missense . Studies of 101 46 ance but also in neuroinflammation that damages neurons 47 TREM2 deficiency in various models of neurological 102 [1]. As such, it is thought that microglia may have a dual 48 disease, such as cuprizone-induced demyelination [6], 103 49 role in AD, becoming dysfunctional over the course of the converge on impaired microgliosis in response to insult, 104 105 50 suggesting that TREM2 may be crucial for microglial activa- 51 106 52 tion. Indeed, genetic targeting of TREM2 in murine AD 107 53 The authors have declared that no conflict of interest exists. models has shown a defective capacity of TREM2- 108 54 1Current address: Eli Lilly and Company, Lilly Corporate Center, 893 S deficient microglia to cluster around Ab plaques. However, 109 Delaware St, Indianapolis, IN 46225, USA. mixed results were reported with regard to plaque burden Q1 *Corresponding author. **Corresponding author. [7,8], and therefore, as with microglia in general, it is not E-mail address: [email protected] (R.T.), mcolonna@ settled whether TREM2 plays a beneficial or detrimental pathology.wustl.edu (M.C.) role in the context of AD.

http://dx.doi.org/10.1016/j.jalz.2016.07.004 1552-5260/Ó 2016 Published by Elsevier Inc. on behalf of the Alzheimer’s Association.

FLA 5.4.0 DTD JALZ2261_proof 9 August 2016 12:13 pm ce 2 W. Song et al. / Alzheimer’s& Dementia - (2016) 1-7

110 To shed light on this question, we sought to establish statistically significant for increased risk in ADSP 177 111 whether TREM2 polymorphisms that increase AD risk are (P 5 .025). Finally, H157Y was found in 5/8 affected indi- 178 112 typically loss-of-function, neutral, or gain-of-function. viduals, but no unaffected individuals were available for 179 113 180 114 Although R47H has been shown to have ligand-binding de- comparison; however, it significantly increased risk in 181 115 fects [7], other AD-associated TREM2 variants have not ADSP (P 5 .01; OR 5 4.7). Thus, we were able to demon- 182 116 been evaluated. To address whether ligand-binding defi- strate increased AD risk for variants other than the previ- 183 117 ciency is a consistent feature of risk alleles, we sought to ously implicated R47H, R62H, and L211P. 184 118 characterize a large panel of mutants. We selected risk al- 185 119 186 120 leles based on whole genome/exome sequencing data and 2.2. Functional characterization of selected TREM2 187 121 tested their function by ligand-binding assays. We find that variants 188 122 most AD-associated variants demonstrate decreased 189 123 signaling in response to a variety of ligands, but surprisingly, Next, we attempted to correlate risk to functional alter- 190 124 some markedly increase signaling. Thus, our results suggest ations. TREM2 binds polyanionic ligands such as phospho- 191 125 192 126 that TREM2 signaling helps protect against AD but can lipids and via the associated intracellular DAP12 193 127 cause harm in excess. Finally, to clarify previous findings leads to downstream protein tyrosine kinase and calcium 194 128 regarding the role of TREM2 conducted in peripheral blood signaling, among other intracellular events [7]. We reasoned 195 129 mononuclear cells (PBMCs) [9,10], we perform a thorough that the multitude of different signals generated by TREM2 196 130 characterization of TREM2 expression in these cells and find in response to receptor ligation are all related to the activa- 197 131 1 198 132 that CD14 human monocytes do not express TREM2 on tion state of the receptor, for which downstream calcium 199 133 the mRNA or protein level, in contrast to monocyte- signaling would provide a reliable readout. Thus, we looked 200 134 derived dendritic cells (moDCs), which highly express for altered TREM2 activation using a calcium-driven re- 201 135 TREM2 and are therefore suitable for functional studies. porter system. Using retroviral transduction, we stably in- 202 136 Our results support the idea that proper TREM2 function serted each TREM2 variant, together with DAP12, into a 203 137 204 138 is important to counteract disease progression. reporter cell line where calcium signals lead to NFAT nu- 205 139 clear translocation and EGFP synthesis. Each reporter line 206 1 140 was sorted directly for the TREM2 population by 207 141 2. Results fluorescence-activated cell sorting (FACS), and these sorted 208 142 lines were used for all subsequent experiments. We charac- 209 143 2.1. Risk analysis of TREM2 variants in NIMH and ADSP 210 144 genetic data terized R47H, R62H, R62C, T96K, L211P, H157Y, and 211 145 D87N from the NIMH families and the previously 212 146 We set out to select TREM2 variants for which we could identified R52H, E151K, R136W, and T66M mutations [12]. 213 147 make inferences about Alzheimer’s risk. We analyzed vari- The common variant of TREM2 (TREM2-CV) and non- 214 148 ants present by whole-genome sequencing in the NIMH transduced reporter cells served as positive and negative 215 149 216 150 AD Genetics Initiative Study and used data on these variants controls, respectively. mRNA expression of TREM2 and 217 151 in the AD Sequencing Project whole-exome sequencing DAP12 by quantitative RT-PCR was similar between con- 218 152 (ADSP) for replication (Table 1). Family-based association structs and highly correlated (Supplementary Fig. 1A), indi- 219 153 analysis (FBAT) in the NIMH AD families yielded a P value cating successful viral transduction. By flow cytometry, 220 154 of .004 for R47H, consistent with our previous report in the surface expression ranged from about 50% to 120% of 221 155 222 156 same NIMH families [11] and confirmed in ADSP TREM2-CV (Supplementary Fig. 1B). The T66M variant 223 157 (P , 3.45e212; odds ratio [OR] 5 4.5). Other rare variants had no detectable surface expression, as previously shown 224 158 in TREM2 were not present in more than ten statistically [13] and did not show any activation in our assay (data not 225 159 informative families, thereby preventing the computation shown). It also demonstrated lower RNA expression, likely 226 160 of reliable P values. However, several variants were frequent because the nontransduced population could not be sorted 227 161 228 162 enough to be suggestive between both data sets. R62H had out by FACS. On the other hand, the R52H, R62C, and 229 163 an OR of 1.7 in NIMH families, confirmed in ADSP T96K mutations demonstrated somewhat lower surface 230 164 (P 5 .006; OR 5 1.4). The related R62C was found in two expression despite their higher RNA expression, which 231 165 affected individuals from one NIMH family and had an may reflect a true defect in protein trafficking or simply tech- 232 166 OR of 0.58 in ADSP, but numbers were too small to draw nical variation. All variants besides T66M demonstrated 233 167 234 168 conclusions. T96K and L211P, which were almost activation when stimulated with plate-bound anti-TREM2 235 169 completely linked in both data sets, had OR around 10 in antibody (Supplementary Fig. 1C), suggesting that these var- 236 170 NIMH families, but this was not supported by the replication iants do not affect the overall folding and conformation of 237 171 sample (P 5 .85; OR 5 1.04; P 5 .59; OR 5 1.13, respec- TREM2. Activation in response to direct antibody- 238 172 tively). On the other hand, D87N had an OR of 0.89 in mediated cross-linkage correlated roughly with surface 239 173 240 174 NIMH families but was significantly associated with risk expression as expected. 241 175 in ADSP (P 5 .017; OR 5 2.3). The stop gain Q33X was Using these reporter lines, we analyzed TREM2 activa- 242 176 biased toward affected individuals in NIMH families and tion in response to a variety of known lipid ligands. Each 243

FLA 5.4.0 DTD JALZ2261_proof 9 August 2016 12:13 pm ce W. Song et al. / Alzheimer’s& Dementia - (2016) 1-7 3

244 Table 1 311 245 TREM2 variants found in the NIMH family-based WGS data and tested for replication in the ADSP case-control samples 312 246 313 Codon #Aff Med. #Unaff Med. NIMH ADSP-WES P value; OR (CI); 247 314 dbSNP rs ID change #Fam-ilies Carr AAO Carr Age MAF OR (P value) MAF 248 315 249 rs201258314 R62C 1 2/2 78 0/0 — 0.00068 NA — .65; 0.58 (0.05–6.43); 0.0001 316 250 rs104894002 Q33* 2 4/4 71 1/3 73 0.00170 NA — .025; (NA); 0.0001 317 251 rs75932628 R47H 20 35/53 72.5 2/14 72 0.01390 11.7 .004 3.45E212; 4.5 (2.7–7.4); 0.005 318 252 rs2234253 T96K 9 11/16 71 1/6 79 0.00475 11 — .85; 1.04 (0.65–1.67); 0.003 319 253 rs143332484 R62H 11 12/23 73 7/18 72 0.00746 1.7 — .006; 1.44 (1.1–1.88); 0.01 320 254 rs2234255 H157Y 4 5/8 74 0/0 — 0.00203 NA — .01; 4.7 (1.04–21.33); 0.0006 321 255 rs2234256 L211P 10 12/18 71 1/6 79 0.00509 10 — .59; 1.13 (0.71–1.81); 0.003 322 256 rs142232675 D87N 5 4/13 76 2/6 76 0.002 0.89 — .017; 2.3 (1.1–5.1); 0.0015 323 257 324 NOTE. #Aff Carr: affected carriers over total affected subjects. Med. AAO: median age of onset in affected carriers. 258 325 NOTE. #Unaff Carr: unaffected carriers over total unaffected subjects. Med. age: median of last known ages of the unaffected carriers. 259 326 Abbreviations: MAF, minor allele frequency; CI, confidence interval. 260 327 261 328 262 ligand was coated on a 96–well plate at different 2.3. Determination of TREM2 expression on human 329 263 330 264 concentrations, and reporter cells containing each TREM2 peripheral blood monocytes and monocyte-derived 331 265 variant were subsequently plated in duplicate. After dendritic cells 332 266 12 hours, the percentage of GFP1 cells at each concentra- 333 267 tion was determined by flow cytometry. Then, the baseline The last part of our study was to determine the utility of 334 268 activation with no ligand was subtracted from each respec- PBMCs for TREM2 studies. Early work in healthy donors de- 335 269 336 tive activation curve (Supplementary Fig. 1D) and area un- tected TREM2 only in tissue macrophages and in vitro 270 cultured dendritic cells and macrophages but not in appre- 337 271 der the curve (AUC) relative to TREM2-CV was used to 338 ciable quantities in blood monocytes [17,18]. However, 272 compare overall activation. An example of a full activation 339 273 curve is shown in Supplementary Fig. 1E. Because of the recent studies reported increased TREM2 expression on 340 274 correlation between surface expression and activation seen blood monocytes from some AD patients [9] and carriers of 341 275 342 with antibody stimulation, we used the activation/expression a CD33 risk allele for AD [10].Wesoughttoclarifythedistri- 276 bution of TREM2 in blood-derived cell populations at the 343 277 ratio to determine functional impact. Fig. 1A–1C show acti- 344 RNA and protein level. Freshly isolated human PBMCs 278 vation versus surface expression plots for the purified lipid 345 279 from normal donors were sorted by magnetic-activated cell 346 ligands phosphatidylserine (PS), sulfatide (Sulf), and phos- 1 2 280 phatidylcholine (PC). As previously shown [7], the R47H sorting into the CD14 monocyte fraction and CD14 frac- 347 281 348 polymorphism has a profound negative impact on signaling tion. Flow cytometry was performed using three different 282 monoclonal antibodies against TREM2 as well as CD14 and 349 283 in response to all tested ligands except PC, which in turn eli- 350 CD33 to verify the sort purity. Although virtually all sorted 284 cited a normal response from all variants. The R62C poly- 1 1 351 285 cells were CD14 CD33 , none of the TREM2 antibodies 352 morphism showed a similarly dramatic reduction in 2 286 activation. The H157Y and R62H variants demonstrated a stained cells to an appreciable degree. The CD14 fraction, 353 287 1 354 lesser defect. To the contrary, T96K and D87N had consis- as expected, did not contain a TREM2 population. However, 288 culturing CD141 cells with IL-4 and GM-CSF for 5 days to 355 289 tently higher activation. R52H, R136W, L211P, and 356 produce moDCs led to de novo expression of TREM2 as de- 290 E151K were neutral overall. The percentage deviations of 357 291 each variant from TREM2-CV across ligands are summa- tected by all three monoclonal antibodies (Fig. 2A). To 358 292 rized in Fig. 1F. As Q33X leads to truncation of nearly the confirm this finding, we performed quantitative RT-PCR on 359 293 1 360 entire protein, it presumably leads to complete loss of func- day0CD14 PBMCs and day 5 moDCs (Fig. 2B). Consistent 294 with our flow cytometry data, TREM2 transcripts of either iso- 361 295 tion and thus was not assayed. 362 form 1 or2 are not present in day 0 human CD14 1 monocytes, 296 As lipoproteins contain a variety of phospholipids, we 363 297 tested whether abundant serum lipoproteins such as high- whereas they are abundant in moDCs. 364 298 density lipoprotein (HDL) and low-density lipoprotein For further support, we analyzed microarray data from 365 299 366 300 (LDL) could activate TREM2 in our reporter system. Indeed, the ImmVar project, including data on peripheral blood 367 301 both HDL and LDL activated TREM2 variants in a similar monocytes from 401 healthy subjects, unstimulated moDCs 368 302 pattern as seen for purified phospholipids (Fig. 1D and 1E). from 30 healthy subjects, and CD41CD62 L1 T lympho- 369 303 These results extend the range of TREM2 ligands and are cytes from 407 healthy subjects [19,20] (Fig. 2C). We found 370 304 consistent with the recent observation that lipoprotein parti- that the log 2 quantile-normalized value for TREM2 tran- 371 305 372 306Q2 cles containing the apolipoprotein E (ApoE)alsobind script in monocytes was consistently low (median 6.33), 373 307Q4 TREM2 [14,15], which is notable given the association of on par with transcripts specific to natural killer cells 374 308 ApoE polymorphisms with AD [16]. (NCR1, median 6.86) and B lymphocytes (CD20, median 375 309 376 310 377

FLA 5.4.0 DTD JALZ2261_proof 9 August 2016 12:13 pm ce 444 443 442 441 440 439 438 437 436 435 434 433 432 431 430 429 428 427 426 425 424 423 422 421 420 419 418 417 416 415 414 413 412 411 410 409 408 407 406 405 404 403 402 401 400 399 398 397 396 395 394 393 392 391 390 389 388 387 386 385 384 383 382 381 380 379 378

web 4C=FPO 4 AD.Rslsaeaeae rmtodnr.()Aayi flremcora aast o ua ooye,mDs n ypoye,showing lymphocytes, T and moDCs, and monocytes, isoforms human TREM2 for of individually. RT-PCR sets plotted quantitative outliers data for with microarray populations shown, large are cell intervals of same 5%–95% Analysis these beta-actin. (C) and from donors. HLA-DRA, isolated an- NCR1, two monoclonal was CD20, different RNA from TREM2, three (B) averaged with donors. staining are for three Results shown for GAPDH. are CD14 obtained histograms 0 Representative were day (A) on results (left) monocytes. 10B11 Similar not and but (left), moDCs 29E3 (right), human TREM2—21E10 by against expressed tibodies is TREM2 2. Fig. L (E) stat and with experiments, HDL, independent (D) three from phosphatidylcholine, data (C) pooled show sulfatide, Plots (B) (F). phosphatidylserine, in summarized (A) significance is for TREM2-CV from shown Hypomorphic activation/expression are in TREM2-CVAUC. difference plots to % normalized Separate mean respectively. subsequently and red, than variant and greater %GFP each significantly blue (activation/expression (A–E) for variants AUC alterations. colored hypermorphic functional obtain and reveals TREM2-CV) to than ligands summed less different concentrations, significantly with (activation/expression of variants stimulated range variants a TREM2 over of measured plots were expression cells surface versus Activation 1. Fig. P , 0 acltdb omSdksmlil oprsn test. comparisons multiple Holm-Sidak’s by calculated .05 L .. DTD 5.4.0 FLA .Sn ta./Azemrs&Dementia & Alzheimer’s / al. et Song W. AZ21pof9Ags 061:3p ce pm 12:13 2016 August 9 JALZ2261_proof 1 ooye rdy5mDs ryhsormrpeet stp control. isotype represents histogram Gray moDCs. 5 day or monocytes - 21)1-7 (2016) RM-V are TREM2-CV) 1 L The DL. reporter istical 511 510 509 508 507 506 505 504 503 502 501 500 499 498 497 496 495 494 493 492 491 490 489 488 487 486 485 484 483 482 481 480 479 478 477 476 475 474 473 472 471 470 469 468 467 466 465 464 463 462 461 460 459 458 457 456 455 454 453 452 451 450 449 448 447 446 445 W. Song et al. / Alzheimer’s& Dementia - (2016) 1-7 5

512 5.05) and similar to the level of TREM2 detected in T lym- On the other hand, two variants we tested (T96K, D87N) 579 513 phocytes (median 5.29). In contrast, TREM2 was expressed showed increased activation across all ligands except for 580 514 highly in moDCs (median, 10.87). We also checked levels of phosphatidylcholine. The D87N variant, in particular, 581 515 582 516 HLA-DRA and beta-actin as controls. HLA-DRA was ex- showed activation/expression ratio in some cases two or 583 517 pressed highly in monocytes and moDCs but not in T lym- three times as high as TREM2-CV. Despite some conflicting 584 518 phocytes (median, 13.07, 13.95, and 6.24, respectively), data between data sets, overall both of these activating mu- 585 519 and beta-actin was expressed highly in all three cell types tations seem to increase AD risk. Although the T96K and 586 520 (median 13.60, 13.85, and 13.47, respectively), as expected. highly linked L211P mutations were not associated with 587 521 588 522 Together, these results indicate that TREM2, on the mRNA increased risk in the ADSP data, their OR in the NIMH 589 523 or protein level, is not appreciably expressed in human data was around 10. In addition, the L211P variant was pre- 590 524 monocytes but at a high level in moDCs. viously found to be significant for risk in African Americans 591 525 [21]. While in that study, the T96 position was not geno- 592 526 typed, it is likely that the observed risk for L211P would 593 527 3. Discussion 594 528 extend to T96K due to their linkage disequilibrium. The 595 529 In our study, we report a novel genetic and functional D87N variant was previously found to be statistically signif- 596 530 analysis of several TREM2 variants. Of these variants, icant for AD risk [2], and we report a similar finding in the 597 531 some have been previously described to be statistically sig- ADSP data set, although this risk is not observed in the 598 532 nificant for risk in various populations, others have been re- NIMH family data. This finding suggests that overactive 599 533 600 534 ported without adequate sample sizes for accurate TREM2 signaling may also promote the development of 601 535 assessment of risk, and some have not previously been re- AD. 602 536 ported. The functional impact of most of these variants, other It remains to be assessed whether differences in ligand 603 537 than Nasu-Hakola variants and R47H, has not been previ- binding and signaling we report in this study directly corre- 604 538 ously determined. We observed that five variants (R47H, spond to microglial activation or function in vivo. It is un- 605 539 606 540 R62H, R62C, H157Y, and R52H) had functional features known which TREM2 ligand or ligands are the most 607 541 indicating impaired TREM2 function. R47H and R62C important during the course of AD, and it could be the 608 542 showed the greatest defect in activation/expression ratio, case that the physiological ligand does not show the same 609 543 whereas R62H and H157Y showed a more modest defect. pattern of deficient binding among TREM2 variants. 610 544 R62C and R52H showed diminished surface expression Another possibility is that different levels of TREM2 activa- 611 545 612 546 despite higher mRNA expression. This latter observation tion may differentially trigger downstream signaling events 613 547 may reflect a true defect in protein trafficking or alterna- in vivo. For example, gross overactivation as seen with T96K 614 548 tively, either technical variation or a deficiency unique to and D87N may end-up recruiting more phosphatases and 615 549 our reporter line rather than human microglia. The R62C leading to decreased microglial activation, cause TREM2 616 550 and R52H variants have not been identified in large enough downregulation, or interfere with other microglial signaling 617 551 618 552 numbers thus far to draw conclusions about AD risk, but the pathways. It should be noted that the L211P variant creates a 619 553 other three showed statistically significant risk in our anal- P-X-X-P motif which may bind SH3 domains on various 620 554 ysis as well as prior reports. The R47H variant significantly ; rather than AD risk being mediated by the linked 621 555 increased AD risk in both the NIMH families and ADSP T96K polymorphism, it could be that risk is being driven 622 556 data, and its association with AD has been replicated in a va- by alterations in the signaling apparatus that cannot be de- 623 557 624 558 riety of populations. The R62H variant is more common than tected by our reporter assay. On the other hand, the D87N 625 559 R47H in the Caucasian population and has previously been variant, which does not have this complicating feature, 626 560 shown to confer a statistically significant risk for AD with also shows increased risk. Further studies in vivo are 627 561 a smaller effect size than R47H, a conclusion supported by required to adequately address the effect of increasing or 628 562 both the NIMH and ADSP data sets. The H157Y variant decreasing TREM2 activation, but our data provide evidence 629 563 630 564 was previously reported in affected individuals, and it con- that changes in the upstream binding event are significant in 631 565 fers a statistically significant increased AD risk in the the context of AD pathogenesis. 632 566 ADSP data set. Thus, we have shown that many different Notably, we observed that phosphatidylcholine led to a 633 567 AD-associated TREM2 risk alleles impair the ability of relatively preserved activation/expression ratio for variants 634 568 the receptor to engage ligand and initiate downstream that significantly enhanced or inhibited responses to other li- 635 569 636 570 signaling, supporting an important role for TREM2 and mi- gands. This discrepancy suggests that there may be multiple 637 571 croglial activation in protecting against AD pathogenesis. binding modes for TREM2 ligands that could be important 638 572 Finally, the Q33X variant together with other Nasu-Hakola in different situations. Additionally, we find that both HDL 639 573 variants that presumably have gross misfolding and loss of and LDL can activate TREM2 in our reporter system, which 640 574 function were previously demonstrated to increase AD risk extends the range of TREM2 ligands and is consistent with 641 575 642 576 in aggregate, and we confirm statistically significant risk the recent observation that lipoprotein particles containing 643 577 for the Q33X variant alone, as well as lack of cell surface apolipoprotein E (ApoE) also bind TREM2 [14,15]. 644 578 expression of the T66M variant. Polymorphisms in apolipoprotein E, which greatly 645

FLA 5.4.0 DTD JALZ2261_proof 9 August 2016 12:13 pm ce 6 W. Song et al. / Alzheimer’s& Dementia - (2016) 1-7

646 modulate AD risk [16], have not been shown to alter TREM2 chloroform:methanol:water 2:1:0.1, respectively. For each 713 647 binding, but the importance of ApoE-TREM2 interaction in experiment, two Nunc-Immuno MicroWell 96 well solid 714 648 AD pathogenesis is yet to be determined. Activation of plates (Sigma) were coated with the purified lipid ligands 715 649 716 650 TREM2 by HDL or LDL may also be important in the pe- phosphatidylcholine, phosphatidylserine, or sulfatide by 717 651 riphery, where many tissue macrophage lineages, including diluting to the indicated concentration in methanol and add- 718 652 arterial wall macrophages, express TREM2. ing 50 mL of the resulting solution to each well, with each 719 653 We demonstrate that peripheral blood monocytes do not condition performed in duplicate. Plates were allowed to 720 654 express TREM2 at a comparable level to cell types, such dry by evaporation, leaving the ligands coated on the well 721 655 722 656 as moDCs, where TREM2 is known to have functional ef- bottom. For human HDL (Millipore) and human LDL (Milli- 723 657 fects. Previous studies that have reported differences in pore), stock solution was diluted to the proper concentration 724 658 monocyte TREM2 mRNA expression [9] or surface expres- in carbonate buffer (15 mM Na2CO3, 35 mM NaHCO3, pH 725 659 sion [10] between different groups of subjects show in- 9.6), and 50 mL of the resulting solution was added to each 726 660 creases on the order of 15% or 40%, respectively. In well, with each condition performed in duplicate. The plates 727 661  728 662 contrast, our data indicate that the increase in TREM2 were transferred to 4 C overnight. The next day, solution 729 663 expression on differentiation of monocytes into moDCs is was aspirated from each well, and each well was washed 730 664 on the order of a hundred to a thousand fold. Thus, absolute once with 150 mL of PBS. For plate-bound antibody, goat 731 665 levels of TREM2 expression in monocytes are quite low and (Fab’)2 anti-mouse IgG (Southern Biotech, cat no. 1012- 732 666 unlikely to be useful for drawing mechanistic conclusions 01) was diluted 1:100 in carbonate buffer, and 50 mLof 733 667 734 668 about TREM2. We believe that future studies of this sort the resulting solution was added to wells in triplicate. The 735  669 should focus on moDCs rather than monocytes. It is possible plate was placed at 4 C for 12 hours. Then, solution was 736 670 that monocyte TREM2 expression may be modulated by aspirated, and each well was washed once with 150 mLof 737 671 some AD-associated systemic signal and thus useful as a PBS. 29E3 hybridoma supernatant was plated in each well 738 672 biomarker, but this remains to be convincingly demon- at 4C. After 12 hours, the supernatant was aspirated and 739 673 m 740 674 strated. wells were washed once with 150 L of PBS. 741 675 Our work provides a link between AD risk and TREM2 After preparation of the plate, 50,000 cells in 75 mLof 742 676 function and provides a useful guide for future human complete media were added to each well. After 12 hours, 743 677 TREM2 studies. the cells were transferred to FACS tube and read on a FACS- 744 678 Calibur. Cells were gated based on forward and side scatter 745 679 746 680 4. Methods and propidium iodide negativity. The nontransduced control 747 681 was used to draw the gate for GFP positivity, and this gate 748 682 4.1. Standard protocol approvals, registrations, and was used to determine the %GFP1 for each variant and 749 683 patient consents ligand concentration. The average %GFP1 at the lowest 750 684 two concentrations for each variant were assumed to be 751 685 Diagnosis of AD dementia was established according to 752 baseline activation, and this value was subtracted from the 686 NINCDA-ADRDA criteria. Informed consent was provided 753 687 curve for each respective variant. The values for all concen- 754 by all participants, and research approval was established by 688 trations were summed to approximate the area under the 755 the relevant institutional review boards in the study cohort. 689 curve, which was used for analysis. 756 690 757 691 4.2. Family-based association testing 758 692 Acknowledgments 759 693 Association analyses of the TREM2 variants in the 760 694 Y.W. was supported by the Lilly Innovation Fellowship 761 NIMH-WGS data set were performed using the family- 695 Award (Eli Lilly and Company). This work was funded by 762 based association test (FBAT as implemented in PBAT 696 grants from NIH 5T32CA009547-30 (T.K.U.), NIMH 763 697 v3.6). See Supplementary Methods for more detail. 764 698 5R01MH060009-14, the JPB Foundation (R.T.), 765 699 RG4687A1/1 from National Multiple Sclerosis Society 766 4.3. Logistical regression in case-control whole-exome 700 (M.Ce.), NIH RF1 AG05148501, and the Cure Alzheimer’s Q3 767 701 data set Fund (M.Co.). 768 702 Authors’ contributions: M.Co. and R.T. designed the study. 769 703 Logistical regression using the binary affection status was 770 704 performed using PLINK v1.90. See Supplementary Methods W.S., S.C.J., and Y.W. conducted experiments. T.K.U. and 771 705 for more detail. M.Ce. provided discussion. B.H. and K.M. analyzed genetic 772 706 data. W.S., B.H., and M.C. wrote the article. 773 707 774 708 4.4. Reporter assay 775 709 Supplementary data 776 710 Phosphatidylcholine (Avanti, #840051P), phosphatidyl- 777 711 serine (Avanti, #840032P), and sulfatides (Avanti, Supplementary data related to this article can be found at 778 712 #131305P) were reconstituted in methanol, methanol, and http://dx.doi.org/10.1016/j.jalz.2016.07.004. 779

FLA 5.4.0 DTD JALZ2261_proof 9 August 2016 12:13 pm ce W. Song et al. / Alzheimer’s& Dementia - (2016) 1-7 7

780 References [12] Jin SC, Benitez BA, Karch CM, Cooper B, Skorupa T, Carrell D, et al. 847 781 Coding variants in TREM2 increase risk for Alzheimer’s disease. Hum 848 782 [1] Heneka MT, Golenbock DT, Latz E. Innate immunity in Alzheimer’s Mol Genet 2014;23:5838–46. 849 783 disease. Nat Immunol 2015;16:229–36. [13] Kleinberger G, Yamanishi Y, Suarez-Calvet M, Czirr E, Lohmann E, 850 784 [2] Guerreiro R, Wojtas A, Bras J, Carrasquillo M, Rogaeva E, Cuyvers E, et al. TREM2 mutations implicated in 851 785 Majounie E, et al. TREM2 variants in Alzheimer’s disease. N Engl J impair cell surface transport and phagocytosis. Sci Transl Med 2014; 852 786 Med 2013;368:117–27. 6:243ra286. 853 787 [3] Jonsson T, Stefansson H, Steinberg S, Jonsdottir I, Jonsson PV, [14] Bailey CC, DeVaux LB, Farzan M. The Triggering Receptor Ex- 854 788 Snaedal J, et al. Variant of TREM2 associated with the risk of Alz- pressed on Myeloid Cells 2 Binds Apolipoprotein E. J Biol Chem 855 789 heimer’s disease. N Engl J Med 2013;368:107–16. 2015;290:26033–42. 856 790 [4] Bertram L, Lange C, Mullin K, Parkinson M, Hsiao M, Hogan MF, [15] Atagi Y, Liu CC, Painter MM, Chen XF, Verbeeck C, Zheng H, 857 791 et al. Genome-wide association analysis reveals putative Alzheimer’s et al. Apolipoprotein E Is a Ligand for Triggering Receptor Ex- 858 792 disease susceptibility loci in addition to APOE. Am J Hum Genet pressed on Myeloid Cells 2 (TREM2). J Biol Chem 2015; 859 793 2008;83:623–32. 290:26043–50. 860 794 [5] Lambert JC, Heath S, Even G, Campion D, Sleegers K, Hiltunen M, [16] Strittmatter WJ, Saunders AM, Schmechel D, Pericak-Vance M, 861 795 et al. Genome-wide association study identifies variants at CLU and Enghild J, Salvesen GS, et al. Apolipoprotein E: high-avidity binding 862 796 CR1 associated with Alzheimer’s disease. Nat Genet 2009;41:1094–9. to beta-amyloid and increased frequency of type 4 allele in late-onset 863 797 [6] Poliani PL, Wang Y, Fontana E, Robinette ML, Yamanishi Y, familial Alzheimer disease. Proc Natl Acad Sci U S A 1993; 864 798 Gilfillan S, et al. TREM2 sustains microglial expansion during aging 90:1977–81. 865 799 and response to demyelination. J Clin Invest 2015;125:2161–70. [17] Cella M, Buonsanti C, Strader C, Kondo T, Salmaggi A, Colonna M. 866 800 [7] Wang Y, Cella M, Mallinson K, Ulrich JD, Young KL, Robinette ML, Impaired differentiation of osteoclasts in TREM-2-deficient individ- 867 801 et al. TREM2 lipid sensing sustains the microglial response in an Alz- uals. J Exp Med 2003;198:645–51. 868 802 heimer’s disease model. Cell 2015;160:1061–71. [18] Bouchon A, Hernandez-Munain C, Cella M, Colonna M. A DAP12- 869 803 [8] Jay TR, Miller CM, Cheng PJ, Graham LC, Bemiller S, Broihier ML, mediated pathway regulates expression of CC chemokine receptor 7 870 804 et al. TREM2 deficiency eliminates TREM21 inflammatory macro- and maturation of human dendritic cells. J Exp Med 2001; 871 805 phages and ameliorates pathology in Alzheimer’s disease mouse 194:1111–22. 872 806 models. J Exp Med 2015;212:287–95. [19] Raj T, Rothamel K, Mostafavi S, Ye C, Lee MN, Replogle JM, et al. 873 807 [9] Hu N, Tan MS, Yu JT, Sun L, Tan L, Wang YL, et al. Increased expres- Polarization of the effects of autoimmune and neurodegenerative 874 808 sion of TREM2 in peripheral blood of Alzheimer’s disease patients. J risk alleles in leukocytes. Science 2014;344:519–23. 875 809 Alzheimers Dis 2014;38:497–501. [20] Lee MN, Ye C, Villani AC, Raj T, Li W, Eisenhaure TM, et al. Com- 876 810 [10] Chan G, White CC, Winn PA, Cimpean M, Replogle JM, Glick LR, mon genetic variants modulate pathogen-sensing responses in human 877 811 et al. CD33 modulates TREM2: convergence of Alzheimer loci. Nat dendritic cells. Science 2014;343:1246980. 878 812 Neurosci 2015;18:1556–8. [21] Jin SC, Carrasquillo MM, Benitez BA, Skorupa T, Carrell D, 879 813 [11] Hooli BV, Parrado AR, Mullin K, Yip WK, Liu T, Roehr JT, et al. The Patel D, et al. TREM2 is associated with increased risk for Alz- 880 814 rare TREM2 R47H variant exerts only a modest effect on Alzheimer heimer’s disease in African Americans. Mol Neurodegener 2015; 881 815 disease risk. Neurology 2014;83:1353–8. 10:19. 882 816 883 817 884 818 885 819 886 820 887 821 888 822 889 823 890 824 891 825 892 826 893 827 894 828 895 829 896 830 897 831 898 832 899 833 900 834 901 835 902 836 903 837 904 838 905 839 906 840 907 841 908 842 909 843 910 844 911 845 912 846 913

FLA 5.4.0 DTD JALZ2261_proof 9 August 2016 12:13 pm ce