Relationships of Alzheimer's disease and apolipoprotein E genotypes with small RNA and protein cargo of brain tissue extracellular vesicles

Alzheimer's disease (AD) is a public health crisis that grows as populations age. Hallmarks of this neurodegenerative disease include aggregation of beta-amyloid peptides and hyperphosphorylated tau proteins in the brain. Variants of the APOE gene are the greatest known risk factors for sporadic AD. As emerging players in AD pathophysiology, extracellular vesicles (EVs) contain proteins, lipids, and RNAs and are involved in disposal of cellular toxins and intercellular communication. AD-related changes in the molecular composition of EVs may contribute to pathophysiology and lend insights into disease mechanisms. We recently adapted a method for separation of brain-derived EVs (bdEVs) from post-mortem tissues. Using this method, we isolated bdEVs from AD patients with different APOE genotypes (n=23) and controls (n=7). bdEVs were counted, sized, and subjected to parallel small RNA sequencing and proteomic analysis. Numerous bdEV-associated RNAs and proteins correlated with AD pathology and APOE genotype. Some of the identified entities have been implicated previously in important AD-related pathways, including amyloid processing, neurodegeneration, and metabolic functions. These findings provide further evidence that bdEVs and their molecular cargo modulate development and progression of AD.

i Department of Surgery, The University of Melbourne, The Royal Melbourne Hospital, Parkville,

INTRODUCTION
tissue debris and spun at 10,000 × g for 30 min at 4°C (Thermo Fisher swinging-bucket rotor 162 model AH-650, k-factor 53, acceleration and deceleration settings of 9, using Ultra-Clear tubes 163 with 5 ml capacity). The pellet was resuspended in 100 μl PBS. This resuspension is termed the 164 "10,000 × g pellet" (10K). The 10,000 × g supernatant was concentrated by 100 kilodalton (kDa) 165 MWCO protein concentrator (Thermo Fisher 88524) from 5 ml to 0.5 ml and loaded onto a size-166 exclusion chromatography column (qEV original, IZON Science SP1-USD, Christchurch, New 167 Zealand) and eluted by PBS. 0.5 ml fractions were collected. The first 3 ml (Fractions 1-6) of the 168 eluate were discarded as the void volume and subsequent 0.5 mL fractions were collected. For 169 the purposes of this study, a total of 2 ml eluate (Fractions 7-10) were pooled and 170 ultracentrifuged for 70 min at 110,000 × g (average) at 4°C (swinging-bucket rotor model TH-171 641, Thermo Fisher, k factor 114 at max speed, acceleration and deceleration settings of 9, 172 using thinwall polypropylene tubes with 13.2 ml capacity). Supernatant was removed, and the 173 pellet was resuspended in 100 μl PBS as the purified EV fraction. Fractions were stored at -80°C. 174 175

Brain homogenate preparation for protein and RNA 176
For protein extraction, brain homogenates (BH) were prepared by grinding tissue in cold PBS 177 containing PI/PS with a handheld homogenizer (Kontes Pellet Pestle Motor) for 10 sec. RIPA onto a trap column (C18 PepMap 100 μm i.d. × 2 cm trapping column, Thermo Fisher Scientific) 249 at 5 µL/min for 6 min using a Thermo Scientific UltiMate 3000 RSLCnano system and washed for 250 6 min before switching the precolumn in line with the analytical column (BEH C18, 1.7 μm, 130 251 Å and 75 μm ID × 25 cm, Waters). Separation of peptides was performed at 45°C, 250 nL/min 252 using a linear ACN gradient of buffer A (water with 0.1% formic acid, 2% ACN) and buffer B 253 (water with 0.1% formic acid, 80% ACN), starting from 2% buffer B to 13% B in 6 min and then 254 to 33% B over 70 min followed by 50% B at 80 min. The gradient was then increased from 50% 255 B to 95% B for 5 min and maintained at 95% B for 1 min. The column was then equilibrated for 256 4 min in water with 0.1% formic acid, 2% ACN. Data were collected on a Q Exactive HF (Thermo 257 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted December 14, 2020. ; https://doi.org/10.1101/2020.12.12.20247890 doi: medRxiv preprint Maxquant V 1.16.0. Trypsin with a maximum of two missed cleavages was used as the cleavage 271 enzyme. Carbamidomethyl of cysteine was set as fixed modification and oxidation of 272 methionine was set as variable modification. The Percolator results were set to reflect a 273 maximum of 1% false discovery rate (FDR). The Label Free quantification was done with match 274 between runs using a match window of 0.7 min. Large label free quantification (LFQ) ratios 275 were stabilized to reduce the sensitivity for outliers. For human datasets, data scaling was done 276 using the cyclic loess method, and scaled data were visualized with a PCA plot. For differential 277 abundance analysis, nested factorial design was set up for the analysis, where each subtype of 278 the disease was nested within the main disease category and contrasts for the main categories 279 were computed by averaging the subtypes. 280

281
The protein interaction and cluster protein function prediction was done by Protein-Protein 282 Interaction Networks Functional Enrichment Analysis (STRING) (82). Kyoto Encyclopedia of 283 Genes and Genomes (KEGG) (83) was used to enrich pathway involvement of identified 284 proteins. Statistical significance of enrichment was determined by the tools mentioned above. 285 Only nominally significant categories (false discovery rate (FDR) < 0.05) were included for 286 analysis. 287 288 Electrochemiluminescence-linked (ECL) immunoassay for total tau and phosphorylated tau 289 detection 290 Total tau and phosphorylated tau at threonine 231 (phosTau) were measured in BH, 10K, and 291 EVs using an ECL-immunoassay (Meso Scale Discovery K15121D) according to the 292 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted December 14, 2020. ; https://doi.org/10.1101/2020.12.12.20247890 doi: medRxiv preprint manufacturer's instructions. In brief, 10K and EV samples were diluted 1:10 while BH was 293 diluted 1:100 in 2% block buffer (2% BSA in tris buffer) containing 0.5% triton X-100. Samples 294 were incubated for one hour on the plate. After washing the plate, the SULFO-TAG Anti-Total 295 Tau Antibody was added and incubated with the plate for one hour. After washing, MSD Read 296 Buffer was added, and the plate was read immediately with a Quick plus SQ 120 MM 297 instrument. Data analysis was done on MSD DISCOVERY WORKBENCH software version 2.0. 298 299 Statistical analysis 300 Statistical significance of particle count, particle: protein ratio, size distribution, RNA biotype 301 differences, and tau and phosTau level differences between AD and control groups and 302 between either of two APOE genotype groups were determined by two-

Separation of EVs from AD and control brain tissue 314
Following the protocol illustrated in Figure 1A, we separated brain-derived extracellular vesicles 315 (bdEVs) from control (n=7) and AD (n=23) individuals with different APOE genotypes (Table 1). 316 A small amount (~50 mg) of each tissue was set aside to produce brain homogenate (BH) to 317 assess protein and RNA profiles of the source material. After enzymatic digestion of the 318 remaining tissue and initial filtering, 10,000 x g ultracentrifuged pellets were collected and 319 termed "10K" as an intermediate product of EV separation. The 10K supernatant was then 320 separated by size exclusion chromatography (SEC) (70) and concentrated into a more pure EV 321 preparation. Fractions were processed for characterization, including proteomics and small RNA 322 profiling. 323 324

Basic EV characterization 325
Particle concentration per 100 mg tissue input and particle size distribution were determined 326 by nano-flow cytometry measurement (NFCM). No significant particle yield difference was 327 detected between AD and control brain-derived 10K and EV fractions ( Figure 1B left). 328 Particle:protein ratio was also calculated to evaluate EV purity. This ratio was similar between 329 AD and control group ( Figure 1B right). However, the EV fraction had a higher particle:protein 330 ratio compared with 10K, as expected and consistent with greater protein contamination of the 331 10K pellet (Figure 1B right). Fractions from AD and control groups also had similar size 332 distributions. However, more small particles in the approximately 45-50 nm diameter range 333 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted December 14, 2020. ; https://doi.org/10.1101/2020.12.12.20247890 doi: medRxiv preprint were observed in AD samples compared with controls ( Figure 1C). No significant particle count 334 differences or size distribution shifts were observed for samples with different APOE genotypes 335 (ε2/3, ε3/3, ε3/4/, and ε4/4; Figure S1A-B). We examined the expression levels of 13 proteins 336 that are commonly reported to be associated with EVs (CD81, CD9, FLOT1, FLOT2, RAB1A,  337   RAB7A, TUBA1B, TUBB4B, ANXA2, ANXA5, ANXA6, ACTN1, GAPDH). These proteins were not 338 found to be significantly differentially abundant between AD and controls. Principal component 339 analysis (PCA) also showed different EV marker expression patterns between BH, 10K, and EV 340 fractions, but not between AD and controls ( Figure 1D). 341 342

Assessment of RNA biotype distribution 343
Small RNA sequencing of BH, 10K, and EV fractions yielded 3.2M (± 1M), 2.2M (± 0.9M), and 344 0.75M (± 0.36M) reads, respectively (M = million, 1 x 10^6). After adapter clipping and 345 removing reads shorter than 15 nt, 71.28% (± 4.4%) of BH, 73.79% (±6.8%) of 10K, and 41.83% 346 (± 9.0%) of EV reads mapped to the human genome (hg38). Reads mapped to various RNA 347 biotypes were normalized to reads per million mapped reads (RPM) (Fig S2A-C). Overall, no 348 major differences in biotype distribution were observed between AD and control conditions, 349 although AD EVs showed a slight decrease in mitochondrial RNA (mtRNA) ( Figure S2C). We then 350 determined enrichment of ncRNA biotypes in EVs relative to tissues and assessed whether the 351 enrichment pattern differed between AD and control brains. To do this, we normalized the RPM 352 of each ncRNA biotype to its average RPM in brain tissues ( Figure S2D). Consistent with earlier 353 reports (70,84-86), miRNA and snoRNA were relatively underrepresented in EVs compared with 354 brain homogenates. The same was observed for mtRNA and, to a lesser extent, for snRNA. Y-355 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint  Figure 2A). Of these, two miRNAs in 10K and nine miRNAs in 367 EVs differed by more than two-fold ( Figure S3A-B). Table 2 provides a summary of these 368 dysregulated miRNAs, indicating up/downregulation and involvement in AD as reported in 369 previous studies, since up/downregulation of some EV-associated miRNAs was consistent with 370 reports of brain tissue expression (54-56). By APOE genotype, focusing on the presumed 371 highest-risk ε4/4 and the lowest-risk ε2/3 carriers in our study, nine miRNAs in 10K differed 372 more than two-fold (with p value < 0.05), while only two miRNAs (miR-379-5p and miR-199a-373 5p) differed in EVs ( Figure 2B, S4A-B, and Table 3). However, none of the same miRNAs were 374 found in the comparisons of AD versus controls and APOE ε4/4 versus ε2/3. 375 376 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted December 14, 2020. ; https://doi.org/10.1101/2020.12.12.20247890 doi: medRxiv preprint To assess whether AD-associated miRNA differences in the extracellular space reflect overall 377 changes in brain tissue, we compared miRNA fold-changes of 10K and EVs with BH. We 378 highlighted miRNAs that were differentially abundant only in BH (blue dots, Figure 2C), only in 379 10K or EV (red dots), or in both EVs and BH (purple dots). Around 30% of differentially abundant 380 EV miRNAs reflected differences that could also be observed in BH (18 miRNAs in 10K and 13 in 381 EV, Figure 2C). miRNAs that showed robust differential abundance (FDR < 0.05 and log2 fold-382 change > 2) were found within this subset of miRNAs. Eight miRNAs showed consistent changes 383 across BH, 10K, and EVs ( Figure 2D). Analyzing the higher-and lower-risk APOE genotypes as 384 above, more than 90 miRNAs were differentially abundant in BH (blue dots), while relatively 385 few miRNAs differed in EVs (red dots). Of the differentially abundant miRNAs, 14 showed 386 consistent change between 10K and BH, but only three between EVs and BH ( Figure 2E). 387 Curiously, the red blood cell-specific miR-451a was consistently differentially abundant across 388 all fractions ( Figure 2F). Whether this is a coincidental contaminant or indicative of some 389 biological difference may require further investigation. miR-328-2p, in contrast, was 390 downmodulated in BH and 10K, but upmodulated in the EV fraction ( Figure 2F). 391 392 393

Differential expression of other non-coding RNA biotypes in brain tissue and EVs related to AD 394 pathology 395
Since incorporation into EVs of non-miRNA non-coding RNAs can be modulated by external 396 stimuli imposed on cells (86,88), and since these ncRNAs may also contribute to intercellular 397 communication (89,90), we examined individual snoRNAs, snRNAs, Y-RNAs, and tRNAs. 398 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted December 14, 2020. ; https://doi.org/10.1101/2020.12.12.20247890 doi: medRxiv preprint snoRNAs were relatively depleted from EVs ( Figure S2D), and we did not find any significant 399 differences in snoRNA content of BH, 10K or EVs in AD versus controls (data not shown). 400 Additionally, no differences in snRNA were found for 10K or EVs in AD versus controls, although 401 a significant upregulation of RNU4-2 in AD brain homogenates was observed ( Figure S5A). We 402 also observed slight differences in Y-RNA content of AD versus control EVs ( Figure S5B). 403 Compared with non-AD controls, 10K fractions of AD patients had higher levels of Y4-RNA (p 404 value < 0.05), while EVs contained lower levels of Y1-RNA (p value < 0.05). 405 406 Interestingly, we found many differences in tRNA content of EVs in AD versus controls ( Figure  407 3A). Most of these differences were subtle, however, as none of the tRNAs were enriched more 408 than twofold. In EVs, we observed changes in various isodecoders of tRNA-Cys-GCA and tRNA-409 Gly-GCC. Additionally, we observed changes in tRNA-Val, tRNA-Arg, tRNA-Pro, and different 410 tRNA-Ala isoacceptors. In contrast, no clear differences in tRNA content were found in 10K. 411 Next, we compared the tRNA content of 10K and EVs from APOE ε4/4 versus ε/3 carriers 412 ( Figure 3B). The fold-differences in tRNA dependent on APOE genotype were larger than those 413 dependent on AD status, although fewer significant tRNAs were found. Two tRNA-Gly-CCC 414 isodecoders were enriched in 10K from APOE ε2/3 carriers, while no significant differences 415 were observed in EVs (as defined by FDR < 0.05). 416

417
We subsequently compared the differences in EV-associated tRNA with differences in brain 418 homogenates ( Figure 3C). Like the miRNAs, most of the differences were observed only in the 419 brain homogenates, or in EVs, while a minor number of tRNAs differed in both. Comparing 420 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted December 14, 2020. ; https://doi.org/10.1101/2020.12.12.20247890 doi: medRxiv preprint APOE genotype-dependent differences in the tRNA content of EVs and BH, we observed a great 421 difference in BH, or in EVs, but only a small number of tRNAs that differed in both ( Figure 3D  4D and Figure S6D). More proteins were up-or downregulated in AD in the EV fraction ( Figure  441 4D, right) than in 10K ( Figure 4D left) or BH ( Figure S6D). EVs have increased potential to 442 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted December 14, 2020. ; https://doi.org/10.1101/2020.12.12.20247890 doi: medRxiv preprint indicate the difference between AD and controls as compared with 10K and BH. Three proteins 443 (10K) and 11 proteins (EVs) differed by more than two-fold between AD and control ( Figure 4E, 444 S7A-B). Examining proteins that were differentially abundant in EVs by Protein-Protein 445 Interaction Networks Functional Enrichment Analysis (STRING), 18 out of 31 had high protein to 446 protein interaction confidence scores (0.7 on a scale of 0-1), participating in AD-related 447 processes such as neurodegeneration, neurotrophin signaling, oxidative regulation, and 448 metabolic regulation ( Figure 4F). Prominent among these was microtubule-associated protein 449 tau (MAPT), which forms neurofibrillary tangles in AD brains. To validate these results, we next 450 measured the concentration of total tau and tau with phosphorylated threonine 231 (phosTau) 451 in lysed BH, 10K, and EVs by ECL-immunoassay. Normalized to total protein input, both total tau 452 and phosTau were significantly increased in bdEVs of AD versus controls, but not in the 10K 453 fraction ( Figure 5A). In BH, only phosTau was significantly increased in AD compared with 454 controls ( Figure 5A). Results were similar when data were normalized by starting amount of 455 tissue ( Figure S8). We then assessed the possible correlations of tau or phosTau in BH and 10K 456 or EV fractions. No strong correlation was observed for total tau levels (BH vs 10K or EVs; 457 Figure 5B) or for phosTau in BH and 10K. However, phosTau levels in BH and EVs were positively 458 correlated (R=0.48, p=0.007; Figure 5B). 459 460 Proteomics: APOE ε4/4 and APOE ε2/3 461 The analyses above were next restricted to APOE ε4/4 and ε2/3 genotypes (AD samples only). 462 Most proteins were detected in more than 50% of individuals in each group (i.e., n > 3 in each; 463 Figure 6A, Figure S9A). Most proteins were found in common in the ε4/4 and ε2/3 groups 464 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted December 14, 2020. ; https://doi.org/10.1101/2020.12.12.20247890 doi: medRxiv preprint (71.4% in 10K, 70.3% in EVs, and 81.5% in BH). KEGG pathway analyses of APOE ε4/4-"unique" 465 proteins in 10K and EVs corresponded with metabolism-related pathways (carbon, glutathione, 466 and galactose metabolism) ( Figure 6B). KEGG pathway analysis of APOE ε2/3-"unique" proteins 467 in 10K and EVs did not return enriched pathways (data not shown), although numerous 468 pathways were enriched in BH of APOE ε2/3 carriers ( Figure S9B). Moreover, both ε4/4 and 469 ε2/3 contained enriched proteins in BH that were involved in Alzheimer's disease, Huntington's 470 disease, and oxidative phosphorylation ( Figure S9B). 471

472
The fold changes and p values of individual proteins upregulated and downregulated in ε4/4 vs 473 ε2/3 are shown in Figure 6C (10K and EVs) and Figure S9C (BH). The expression levels of 474 dysregulated proteins between ε4/4 vs ε2/3 are shown in Figure 6D and Figure S10A The roles of EVs in regulating CNS diseases have been inferred predominantly from studies of in 483 vitro models and biofluid EVs, with growing but still limited study of tissue EVs. Here, we 484 compared protein and small RNA contents of brain homogenate with those of a "10K" pelleted 485 extracellular fraction and a purified EV fraction of control and late-stage AD brain, including 486 several APOE genotypes. Although the concentrations of recovered particles did not differ 487 significantly with AD pathology, several bdEV miRNAs, tRNAs, and proteins were dysregulated 488 not only between AD and controls, but also between the APOE genotypes representing the 489 most "distant" risk groups in our study, the ε4/4 and ε2/3 carriers. Proteome differences 490 between AD and controls were most pronounced for EVs, suggesting that EV proteins may have 491 the best biomarker potential, at least for late-stage AD. The dysregulated molecules identified 492 in our study, especially those involved in aging and neurodegeneration pathways, may be 493 involved in CNS disease mechanisms and constitute new biomarkers for disease monitoring. 494 495

Does AD or APOE genotype affect EV production? 496
Consistent with previous studies on bdEVs from human cortex (91,92), we observed a similar 497 bdEV recovery for AD and control. However, since bdEVs are released as a mixture of various 498 subtypes and from diverse cells, this finding does not exclude the possibility that biogenesis of 499 specific EV subtypes may be affected by AD. For example, MHC class I bdEVs were reportedly 500 enhanced in preclinical AD patients compared with controls and late-stage AD (92). Also, the 501 APOE ε4 allele was reported to be involved in neuronal, endosomal, and lysosomal system 502 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted December 14, 2020. ; https://doi.org/10.1101/2020.12.12.20247890 doi: medRxiv preprint dysfunction (93). In another study, control brain tissues from human and mouse had reduced 503 EV counts in the presence of the APOE ε4 allele (61). However, in our late-stage AD samples 504 (Braak stage 5-6, CERAD B-C), no particle concentration difference was associated with the 505 APOE ε4 allele or other variants. Consistent with previous findings, this suggests that effects of 506 APOE alleles, if present, may affect EVs and their effects on brain function in a measurable 507 manner only before late-stage disease (13,94). Further studies of APOE will be needed in young 508 brain and during early disease stages. 509 510 Altered bdEV miRNA expression in AD 511 RNA sequencing revealed several differentially expressed (DE) miRNAs in AD bdEVs with 512 potential involvement in AD. Importantly, these DE miRNAs largely overlapped with results of 513 earlier studies on miRNAs in AD brain tissues (54-56). For example, miRNA-132-3p was 514 downregulated in AD EVs. It was previously found to be downregulated in late-onset AD, 515 negatively correlated with Braak stages and formation of tau tangles in the prefrontal cortex 516 (55). miR-132 promotes neurite outgrowth and synapse formation (95,96), with expression 517 driven by the neurotrophin-responsive transcription factor CREB (97). Additionally, miR-132 518 knockout mice express increased levels of phosphorylated tau (98) and display memory deficits 519 (99). In line with downregulated miR-132-5p, we found that tau levels were increased in AD 520 EVs. Two other miRNAs downregulated in AD EVs (miR-485-3p and miR-338-5p) target beta-521 secretase 1 (BACE1) (100,101). Together with γ-secretase, BACE1 cleaves APP into aggregate-522 forming Aβ peptides (102). Downregulated miR-129-5p and miR-432-5p regulate 523 neuroinflammation (103,104), and miR-129-3p targets fragile X mental retardation 1 (FMR1), 524 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted December 14, 2020. ; https://doi.org/10.1101/2020.12.12.20247890 doi: medRxiv preprint which controls migration of cortical neurons (105,106). Although our study predominantly 525 highlights downregulated miRNAs, upregulated miRNAs have also been reported in AD bdEVs 526 (107). We found at least three upregulated miRNAs in common with that study: miR-483-5p, 527 miR-152-3p, and miR-30a-5p (107). Beyond these validations, differences between the studies 528 may be explained by use of different brain regions for bdEV isolation, as well as differences in 529 separation methods, possibly affecting purity and recovery of EV subtypes. Overall, we found 530 various dysregulated miRNAs in AD EVs that appear to be involved in AD hallmarks such as APP 531 and tau processing, neuronal function, memory formation, and neuroinflammation. 532 533 bdEV miRNAs and APOE genotype 534 In contrast with AD versus control comparisons, EV miRNA profile differences by APOE were not 535 as pronounced. Nevertheless, several EV-associated miRNAs differed. In the 10K fraction of 536 APOE ε4/4 carriers, miR-29b-5p was downregulated. This miRNA targets BACE, which 537 contributes to Aβ aggregation (Hebert et al PNAS 2008). Additionally, we observed 538 downregulation of miR-379-5p and miR-410-3p, which have been previously implicated in 539 anxiety behavior (108), miR-23a-5p, which is linked to neuronal cell loss (109), and miR-483-5p, 540 which is involved in neuronal development (110). Our findings are limited, however, since only 541 late-stage carriers were examined, with late Braak stage and similar CNS pathology. miRNAs are 542 quite possibly involved in APOE genotype interactions much earlier in AD, perhaps even in 543 young, putatively healthy individuals. More research is thus needed to address the involvement 544 of these miRNAs in AD pathogenesis. 545 546 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted December 14, 2020. ; https://doi.org/10.1101/2020.12.12.20247890 doi: medRxiv preprint AD-dependent differences in additional bdEV-associated ncRNAs 547 AD-dependent differences in bdEV-associated ncRNAs such as tRNAs and Y-RNAs were also 548 apparent. These ncRNAs are highly abundant in EVs from various sources (84-86,111-113), but 549 their roles in health and disease remain largely unexplored. We found various AD-dependent 550 differences in tRNA content of EVs, as well as APOE-dependent differences in tRNAs in the 10K 551 fraction. Changes in cellular tRNA abundance have been implicated in neurological disorders 552 (114,115), and dysregulation of cellular tRNA levels affects protein expression in cancer (116). 553 Interestingly, tRNA fragments released with EVs may modulate gene expression in recipient 554 cells (117). Small changes in bdEV-associated Y-RNAs (Y1 and Y4) were also observed. Y-RNA is 555 abundantly detected in EVs from various biological sources (118). Cellular Y-RNA serves as a 556 scaffold for RNA-binding proteins (119). In AD patients, dysregulation of Y3-RNA binding to 557 enhancer protein HuD (ELAVL4) was shown to cause alternative splicing in neurons (120). 558 Additionally, EV-associated Y4-RNA activated endosomal RNA sensor TLR7 in monocytes (90). In 559 brain tissue of AD patients, we observed a significant difference in U4 RNA, which has not been 560 previously reported in AD. As part of the spliceosome complex, U4 is involved in pre-mRNA 561 splicing (121). In the prefrontal cortex of AD brains, snRNA U1 was found to be upregulated and 562 to accumulate in tau plaques (122,123). Upregulation of U1 was additionally implicated in 563 aberrant mRNA splicing and altered APP expression (122). Our findings encourage further 564 investigations into the role of extracellular non-miRNA ncRNAs in AD and other neurological 565 conditions. 566 567

Implications of bdEV-associated proteins in AD 568
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted December 14, 2020. ; https://doi.org/10.1101/2020.12.12.20247890 doi: medRxiv preprint

bdEV-associated tau proteins 569
EVs have been widely reported to propagate misfolded protein MAPT (tau) in 570 neurodegenerative disease (42,124-126). Consistent with other bdEV studies (42,91,125), we 571 found a higher level of tau and phosTau protein in EVs of AD brain compared with controls, 572 adding more evidence to EV-mediated tau propagation. However, we did not see consistent 573 associations with APOE genotype, possibly because our samples sizes were small, and possibly 574 because we examined brains from individuals with late-stage disease. Further study is needed, 575 including study of younger and early-stage individuals with different genotypes, to decide if EVs 576 act as a protective factor to clear tau from the CNS to the periphery, or as a neurotoxic factor to 577 spread tau between neurons and glia, or perhaps some as-yet incompletely understood balance 578 between the two extremes. 579 is the first study to show PRDXs upregulating in tissue EVs from AD brain. We also found 589 another antioxidant defense protein, deglucase DJ1 (PARK7), enriched in APOE ε4/4 compared 590 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. proteins, the 14-3-3 protein family of ubiquitous phosphoserine/threonine-binding proteins is 612 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted December 14, 2020. ; https://doi.org/10.1101/2020.12.12.20247890 doi: medRxiv preprint highly abundant in brain and implicated in nervous system development and maintenance 613 (151-153). Our study shows dysregulation of 14-3-3 isoforms ε (YWHAE), ζ (YWHAZ), and γ 614 (YWHAG) in the 10K and EV fractions. 14-3-3 proteins indirectly regulate activation of Rho 615 family GTPases (154). Interestingly, several proteins in the Rho-GTPase cycle were also 616 dysregulated in our bdEVs, including Ras-related C3 botulinum toxin substrate 1/3 (RAC1/3), 617 and Rho GDP-dissociation inhibitors 1 and 2 (ARHGDIA, GDI2). We also found that contactin-1 618 (CNTN1) and neurofascin (NFASC), two proteins involved in axonal guidance and neuron 619 projection development (155-158), were downregulated in AD bdEVs, which may reflect 620 neurodegeneration. In addition, some neuroprotective proteins were also found to be 621 downregulated in bdEVs from APOE ε4/4 patients compared to ε2/3, such as γ-enolase (ENO2) 622 Following digestion, centrifugation, and filtration steps, 10,000 x g pellets from AD and control 661 brain tissue (BH) (as indicated in Table 1) were collected and defined as the 10K fraction. Size-662 exclusion chromatography (SEC) was applied to 10,000 x g supernatants to enrich EVs. RNA and 663 proteins from BH, 10K, and EVs were then isolated and subjected to small RNA sequencing and CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review) preprint
The copyright holder for this this version posted December 14, 2020. ; https://doi.org/10.1101/2020.12.12.20247890 doi: medRxiv preprint  is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted December 14, 2020. ; https://doi.org/10.1101/2020.12.12.20247890 doi: medRxiv preprint dysfunction: An early event in Alzheimer pathology accumulates with age in AD 1261 transgenic mice. Neurobiol Aging. 2009; 1262 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted December 14, 2020. ; https://doi.org/10.1101/2020.12.12.20247890 doi: medRxiv preprint . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted December 14, 2020. ; https://doi.org/10.1101/2020.12.12.20247890 doi: medRxiv preprint      CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetu (which was not certified by peer review) preprint