Selective sorting of microRNAs into exosomes by phase-separated YBX1 condensates

Exosomes may mediate cell-to-cell communication by transporting various proteins and nucleic acids to neighboring cells. Some protein and RNA cargoes are significantly enriched in exosomes. How cells efficiently and selectively sort them into exosomes remains incompletely explored. Previously we reported that YBX1 is required in sorting of miR-223 into exosomes. Here we show that YBX1 undergoes liquid-liquid phase separation (LLPS) in vitro and in cells. YBX1 condensates selectively recruit miR-223 in vitro and into exosomes secreted by cultured cells. Point mutations that inhibit YBX1 phase separation impair the incorporation of YBX1 protein into biomolecular condensates formed in cells, and perturb miR-233 sorting into exosomes. We propose that phase separation-mediated local enrichment of cytosolic RNA binding proteins and their cognate RNAs enables their targeting and packaging by vesicles that bud into multivesicular bodies. This provides a possible mechanism for efficient and selective engulfment of cytosolic proteins and RNAs into intraluminal vesicles which are then secreted as exosomes from cells.

The selectivity for cargo sorting into EVs is best studied for RNA molecules. corresponding to exosomes. To test the presence of YBX1 in EVs from cultured U2OS 241 cells, we examined the fractionation of extracellular YBX1 by differential centrifugation 242 and found by immunoblot that endogenous YBX1 co-sedimented with multiple EV 243 markers (Fig. 4C). Overexpression of YFP-YBX1 in ΔYBX1 cells enhanced the 244 secretion of sedimentable YBX1 (Fig. 4C). Five-fold more YFP-YBX1 was detected in 245 the sediment of culture medium from cells overexpressing wt compared to F85A mutant 246 YBX1 fusion protein. IDR defective YBX1 mutant proteins (CTD-Y to S, CTD-RK to 247 G) were less efficiently packaged than wt YBX1 into extracellular vesicles (Fig. 4D). To 248 confirm that YBX1 resided inside the lumen of extracellular vesicles, we performed 249 proteinase K protection assays on membranes in the high-speed pellet fraction. As Fig.  250 4E shows, endogenous YBX1 was protected from proteinase K digestion in the absence 251 but not in the presence of Triton X-100. ALIX, a cytosolic protein within exosomes, and 252 Flotillin-2, a membrane protein anchored to the inner leaflet of EVs, served as positive 253 controls that were also degraded only in the presence of detergent. CD9, a multi (putative 254 four)-transmembrane protein with an extracellular loop recognized by CD9 antibody, was 255 vulnerable to degradation independent of detergent. Similarly, YFP-tagged YBX1 from a 256 high-speed pellet fraction was mostly resistant to proteinase K (Fig. 4F). These results 257 confirmed that YBX1 was packaged into exosomes secreted from U2OS cells. In 258 contrast, the RNA-binding defective YBX1 mutant F85A and IDR defective mutants (Y 259 to S and RK to G) were significantly decreased in high-speed pellet fractions (Fig. 4C  260 and 4D) 261 Cells overexpressing YFP-YBX1 were used for further purification of EVs by 262 buoyant density flotation (Fig. 4G). Isolated vesicles from U2OS averaged around 130 263 nm in diameter as determined by nanoparticle tracking analysis (NTA) (Fig. 4H, vesicles 264 from HEK293T cells averaged around 100 nm in diameter, Fig. 4 supplement 1B). 265 Vesicles examined by negative stain electron microscopy displayed a characteristic cup-266 shape ( Fig. 4 supplement 1C). YFP-YBX1 was detected in the buoyant vesicle fraction 267 from ΔYBX1/YFP-YBX1 cells but not from ΔYBX1 cells (Fig. 4I). Further separation 268 of these vesicles was achieved on a linear iodixanol gradient (5-25%) which resolved two 269 distinct EV species: low density (LD) and high density (HD) sub-populations, as we 270 previously reported for EVs from MDA-MB-231 cells (Temoche-Diaz et al., 2019) ( Fig.  271 4J and 4K). The YBX1 signal was detected in the combined HD vesicles which 272 coincided with the exosome markers CD63 and ALIX (Fig. 4K). Approximately 10-fold 273 more YFP-YBX1 than YFP-YBX F85A mutant protein was detected in the HD vesicle 274 fractions normalized to CD9 content in each (Fig. 4K). These data illustrate that YBX1 275 sorting into exosomes is dependent on both binding to RNA and IDR-driven phase 276 separation. 277

IDR-driven YBX1 phase separation is required for sorting miRNA into exosomes 279
To test whether YBX1 condensation correlated with exosomal RNA sorting in cells, we 280 reexamined the YBX1-dependent enrichment of miR-223 in exosomes purified by 281 buoyant density flotation, as described in Fig. 4G, from two different cell lines, 282 HEK293T and U2OS. As before, we found that miR-223 and miR-144 were 283 significantly enriched whereas cytoplasmic miR-190 was not enriched in purified 284 exosomes compared to cells (Fig. 5A) (Shurtleff et al. 2016). Overexpression of YBX1 285 increased the relative miR-223 level in exosomes in both cell lines (Fig. 5B). To confirm 286 the requirement of YBX1 in sorting miR-223 into exosomes, we generated a YBX1 287 knockout HEK293T cell line with CRISPR/Cas9. YBX1 knockout clones (ΔYBX1-9, 288 and ΔYBX1-41) were confirmed by Sanger sequencing for target DNA (Fig. 5  289 supplement 1A), RT-qPCR for mRNA ( Fig. 5 supplement 1B) and immunoblot for 290 YBX1 protein (Fig. 5C). A similar knockout was made with U2OS cells (Lyons et al., 291 2016). We used RT-qPCR to quantify miR-223 levels in cells and exosomes, purified as 292 described in Fig. 4G. MiR-223 secretion into the growth medium and in isolated 293 exosomes was reduced ~ 2-fold and correspondingly accumulated within ΔYBX1 mutant 294 derivatives of HEK293T and U2OS cells ( Fig. 5D and Fig. 5 supplement 1C). Finally,295 to test the role of the YBX1 RNA-binding CSD and the CTD in sorting of miR-223 into 296 exosomes, we overexpressed YBX1-F85A, YBX1-Y to S and YBX1-RK to G mutants in 297 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.06.451310 doi: bioRxiv preprint ΔYBX1 U2OS cells and found similar miR-223 reductions in exosomes and 298 accumulation within cells in all three mutant lines (Fig. 5E). Similar results were seen 299 for secretion of miR-223 into the medium fraction of ΔYBX1 mutant 293T cells 300 overexpressing CTD mutants of YBX1 (Fig. 5 supplement 1D) increasing the RNA/YBX1 ratio initially promoted liquid droplet size until a point where 309 droplets were less stable or were not produced ( Fig. 6 supplement 1A). 310 Given the cellular function of YBX1 involves sorting miR-223 into exosomes, we 311 examined miR-223 capture into YBX1 droplets. Cy5( 5') labeled miR-223 was 312 incubated with mGFP-YBX1 under phase separation conditions and observed by 313 fluorescence microscopy. As shown in Fig. 6A, miR-223 accumulated in liquid-like 314 droplets coincident with YBX1. 315 We sought to identify the domains of YBX1 that contribute to the recruitment of 316 miR-223. Cy5 (5') labeled miR-223 was mixed with different YBX1 variants as shown 317 in Fig. 6B. Disrupting the association of YBX1 and miRNA through mutation of Phe85 318 to Ala had no effect on YBX1 droplet formation, but almost completely blocked miR-223 319 recruitment ( Fig. 6B and 6C). The YBX1-CTD-Y to S mutant greatly reduced the 320 formation of droplets but those that formed recruited miR-223 at a similarly reduced level 321 ( Fig. 6B and 6C). miR-223 condensation was not detected when YBX1 phase separation 322 was completely blocked in the YBX1-CTD-RK to G mutant ( Fig. 6B and 6C). These 323 data suggest that YBX1 recruits miR-223 through direct interaction with the central, cold 324 shock domain, and into condensates governed by the C-terminal domain. 325 326 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.06.451310 doi: bioRxiv preprint

YBX1 condensates recruit miRNAs and sort them into exosomes with selectivity 327
To examine whether YBX1 recruits miRNAs with selectivity, we analyzed two 328 additional miRNAs: miR-190, an abundant cellular miRNA and miR-144, one that is 329 highly enriched in exosomes (Shurtleff et al., 2016). We first incubated 5'Cy5 labeled 330 miR-223, miR-190 and miR-144 with mGFP-YBX1 independently. Incubations were 331 conducted in the presence of an excess of unlabeled RNA (10 ng/ul total RNA was 332 extracted from U2OS cells) which produced a clear discrimination between miR-223 333 which partitioned well into YBX1 condensates from miR-190 and miR-144, which did 334 not ( Fig. 6D and 6E). The addition of unlabeled excess RNA produced enlarged YBX1 335 droplets ( Fig. 6 supplement 2A). At a fixed concentration of 7.5uM YBX1 and 100nM 336 miRNA but without unlabeled excess RNA, both miR-223 and miR-190 partitioned into 337 YBX1 droplets (Fig. 6 supplement 2B and 2C). However at varied concentrations of 338 miRNA, the partition coefficient for miR-223 into YBX1 droplets was higher than miR-339 190 while the partition coefficients for YBX1 were almost same ( Fig. 6 supplement 2C). 340 Additional tests of selectivity were conducted in the presence of unlabeled cellular RNA. 341 Unlike miR-190, miR-144 was detected enriched in EVs and dependent on YBX1 342 for secretion from HEK293T cells (Shurtleff et al., 2016). Nonetheless, miR-144 was not 343 recruited in YBX1 droplets ( Fig. 6D and 6E). In an independent assay, we evaluated the 344 interaction of YBX1 with three miRNAs by co-immunoprecipitation from HEK293T cell 345 lysates. These results were consistent with the capture or not of these miRNAs in YBX1 346 condensates with nearly quantitative co-precipitation of YBX1 and miR-223 but 5-10-347 fold lower co-precipitation of miR-190 and miR-144 from cell lysates (Fig. 6F). We next 348 assessed miRNA partition into the phase separation defective mutants YBX1-CTD-Y to 349 S and YBX1-CTD-RK to G (Fig. 6G, 6H and 6I). The Y to S mutant produced less 350 condensate at the same protein concentration and was somewhat less discriminatory and 351 the RK-G mutant produced no visible condensate of protein or RNA. Although the 352 results of these two experiments were consistent, the requirement of YBX1 for secretion 353 of miR-144 in cells was not reflected in a requirement for capture by pure YBX1 protein 354 in condensates. 355 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.06.451310 doi: bioRxiv preprint In our previous work, we evaluated the requirement for YBX1 in the secretion of 356 miR-223 and miR-144 by assaying samples of the culture medium in which HEK293T 357 cells were grown. We sought to refine this measurement by quantifying the miRNA 358 content of buoyant vesicles secreted in the culture medium as we did for the experiment 359 in Fig. 5D. As shown in Fig. 5A, among these three miRNAs, miR-223 and miR-144 but 360 not miR-190 were enriched to different extents in exosomes secreted by HEK293T and 361 U2OS cells. Sorting of miR-223 but not miR-190 and miR-144 into exosomes was 362 decreased in ΔYBX1 cells (Fig. 6J). Combined with the other results of our current work, 363 we conclude, as before, that YBX1 enhances the secretion of miR-223 in EVs but that 364 miR-144, though enriched in EVs, does not engage YBX1 condensates and is not 365 required for secretion in EVs. It seems likely that another RBP is responsible for sorting 366 of miR-144 into exosomes. 367 368

Condensation of YBX1 into P-bodies is required for sorting miRNAs into exosomes 369
YBX1 was previously suggested to be a component of P-bodies that form foci 370 together with Dcp1a as visualized by fluorescence microscopy (Yang and Bloch, 2007). 371 We observed endogenous YBX1 colocalized with P-body components, EDC4, Dcp1a and 372 DDX6, as visualized by specific antibodies (Fig. 7A and 7B). To address the role of 373 condensation in the incorporation of YBX1 in P-bodies, we analyzed the co-localization 374 of IDR mutants with EDC4. Both YBX1-CTD-Y to S and YBX1-CTD-RK to G mutants 375 largely eliminated YBX1 condensation (Fig. 7C). 376 To further study the association of YBX1 and P-bodies, we performed affinity 377 purification coupled with mass spectrometry analysis of N-terminally EGFP-tagged or 378 3xFlag-tagged YBX1. Comparing positive hits found with both tagged forms of YBX1 379 immunoprecipitation trials, we generated a proteome of potential YBX1 interactors and 380 compared this with a published P-body proteome (Hubstenberger et al., 2017) ( Table 1, 381 Fig. 7D). About 35% (43/125) of P-body proteins were identified as potential YBX1 382 binding partners. Gene Ontology analysis showed that RNA-binding proteins were 383 enriched in the YBX1 interactome (Fig. 7E). Some of the RNA-binding proteins, such as 384 SYNCRIP and SSB (Lupus La protein), were identified previously for roles in sorting 385 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 6, 2021. observed that SYNCRIP formed condensates and co-localized with YBX1( Fig. 7  387 supplement 1A) in cells, implying that the condensation properties might be shared by 388 other RBPs that are involved in exosomal RNA sorting. The potential YBX1 interactors 389 included components of the miRNA processing pathway, MOV10 and Ago2, and well-390 known P-body markers, DDX6 and EDC4 (Fig. 7F). immunoprecipitation, we found that wild type but not condensation-defective mutant 396 forms of YBX1 interacted with DDX6 ( Fig. 7G and 7H). Correspondingly, we found 397 that DDX6 was sorted into the luminal interior of isolated EVs as judged by buoyant 398 density fractionation and a proteinase k protection assay (Fig. 7I). To extend this 399 analysis, we purified EVs from HEK293T cells by buoyant density flotation as described 400 in Fig. 4G and conducted a proteomic analysis using liquid chromatography tandem mass 401 spectrometry (LC-MS/MS) ( Table 2). Compared with the published P-body proteome, 402 we found that 18.4% (23/125) of P-body proteins were identified in exosomes ( Fig. 7J  403 and Table 2 was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 6, 2021. was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.06.451310 doi: bioRxiv preprint Of the two IDRs in YBX1, only the one located in the C-terminal domain appears 447 to influence the phase condensation properties of the protein . Within the C-terminal IDR,  448 we identified Y, R and K residues that contribute to LLPS. As shown for prion-like RNA 449 binding properties, we suggest that phase separation of YBX1 is governed by interactions 450 between Y and R residues (Wang et al., 2018). RNA-binding through an interaction with 451 the CSD domain of YBX1 appears to reinforce phase separation producing larger 452 droplets at an optimum ratio of RNA/protein ( Fig.6 supplement 1A). into the endosome. We suggest that molecules destined for secretion in exosomes are 471 segregated in a two-step process involving partition into a precursor larger RNA granule 472 or into RNA granules of distinct function followed by sorting from or among granules to 473 capture those molecules fated for secretion in exosomes and capture by target cells into 474 which exosomes are internalized (Fig. 8). 475 476 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 6, 2021.   Conditioned medium (about 420 ml) was harvested from HEK293T or U2OS cultured 497 cells at 80% confluency. All the following manipulations were performed at 4 C. Cells 498 and large debris were removed by centrifugation at 1500xg for 20 min in a Sorvall R6+ 499 centrifuge (Thermo Fisher Scientific) followed by 10,000xg for 20 min in 500 ml vessels 500 using a fixed angle FIBERlite F14-6 x 500y rotor (Thermo Fisher Scientific). The 501 supernatant fraction was then centrifuged onto a 60% sucrose cushion in buffer A (10 502 mM HEPES pH 7.4, 0.85% w/v NaCl) at ~ 100,000xg (28,000 rpm) for 1.5 h using SW 503 32 Ti swinging-bucket rotors. The interface on the sucrose cushion was collected and 504 pooled from three tubes and applied onto a 60% sucrose cushion for an additional 505 centrifugation at ~ 120,000xg (31,500 rpm) in a SW 41 Ti swinging-bucket rotor for 16 506 h. The sucrose concentration of the collection from the first sucrose cushion interface was 507 measured by refractometry and was adjusted to a concentration <20%. Higher 508 concentrations of sucrose impede sedimentation because EVs equilibrate at a buoyant 509 density above that level. For purification, EV subpopulations that resolve at distinct 510 buoyant densities in a linear gradient were collected and mixed with 60% sucrose to a 511 final volume of 4 ml (sucrose final concentration is ~ 48%). Layers of 1.5 ml of 25%, 512 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.06.451310 doi: bioRxiv preprint 20%, 15%, 10%, and 5% iodixanol (Optiprep) solution in buffer A were sequentially 513 overlaid and samples were centrifugated at ~ 150, 000xg (36,500 rpm) for 16 h in a SW 514 41 Ti rotor. Fractions (400 ul for each) from top to bottom were collected and mixed with 515 SDS sample buffer for immunoblot analysis. In some cases, such as in the immunoblot of 516 YBX1, the floated fractions corresponding to the high density from 1.13 to 1.15 g/ml 517 were pooled and concentrated by centrifugation to improve detection by immunoblot. 518 519 For EV purification in bulk (without discriminating among EV sub-populations), the first 520 cushion-sedimented vesicles above were collected and mixed with 60% sucrose to a final 521 volume of 8 ml. At this point it was important to keep the sucrose concentration > 50%. 522 Aliquots (3 ml) of 40%, (1.5 ml)10% sucrose buffer were sequentially overlaid and the 523 tubes were centrifuged at ~ 150, 000xg (36,500 rpm) for 16 h in a SW 41 Ti swinging-524 bucket rotor. The 10/40% interface was collected and used either directly for RNA 525 extraction by a mirVana miRNA isolation kit (Thermo Fischer Scientific) or washed with 526 PBS and concentrated by centrifugation at ~ 120, 000xg in a SW 55 Ti rotor for 70min. 527 Samples were then prepared for immunoblot analysis. 528 529 For proteinase K protection assays, the supernatant fraction from 10,000xg of 530 conditioned medium was centrifuged at 100,000xg (28,000 rpm) for 1.5 h using SW 32 531 Ti rotors. Pellet fractions resuspended in PBS were pooled and centrifuged at ~ 150, 532 000xg (36,500 rpm) for 70min in a SW 55 Ti rotor. The pellet was resuspended in PBS 533 and split into four equal aliquots. One sample was left untreated, another sample was 534 treated with 0.5% Triton X-100, the third sample was treated with 5 ug/ml proteinase K 535 on ice for 20 min, and the last one was mixed with 0.5% Triton X-100 prior to proteinase 536 K treatment. Proteinase K was inactivated with 5 mM phenylmethane sulfonyl fluoride 537 (PMSF) on ice for 5 min and samples were then mixed with SDS sample loading buffer 538 for immunoblot analysis. 539 Nanoparticle tracking analysis 540 Extracellular vesicles purified by buoyant density centrifugation were diluted 1:100 with 541 PBS filtered with a 0.02 um filter (What GmbH, Dassel, Germany). The liquid was drawn 542 into a 1 ml syringe and inserted into a Nanosight LM10 instrument equipped with a 405-543 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 6, 2021.

Constructs, protein expression and purification 664
Plasmid information is listed in the key resources table. Maltose-binding protein hybrid 665 genes were expressed and the fusion proteins were isolated from baculovirus-infected 666 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 6, 2021.

In vitro phase separation assays 710
For droplet formation without crowding agents (Fig 2A, up panel, Fig 2B, Fig 2C,  The recovery was recorded at a rate of 30 ms/frame, 40 frames in total. For imaging cells, 753 FRAP was performed using an inverted laser scanning confocal microscope (Zeiss, LSM 754 880 AxioImager) equipped with a full incubation chamber maintained at 37 C and 755 supplied with 5% CO2. The point region was bleached with 10 iterations of 100% of 756 maximum laser power of a 514 nm laser. The recovery was recorded at the rate of 2 757 s/interval, 120 cycles in total for YBX1-WT and 1s/interval, 60 cycles in total for YBX1-758 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 6, 2021. BL_corr2 (t) was further normalized to the mean pre-bleach intensities, which were 766 used to estimate 100% fluorescence intensity: 767 Finally, an exponential recovery-like-shape curve was generated by plotting the 769 normalized fluorescent intensity value to times. 770

Live-cell imaging under hexanediol treatments 771
Live cell imaging was performed on an LSM880 microscope with the incubation 772 chamber maintained at 37 C and 5% CO2. U2OS cells expressing YFP-YBX1 were 773 grown on a coverslip-bottom in 35-mm dishes (MatTek P35G-1.5-14-C) until 774 approximately 70% confluency and then imaged using the 514 nm laser. The stock 775 solutions of 1,6-hexanediol (Sigma-Aldrich, 240117) and 2,5-hexanediol (Sigma-Aldrich, 776 H11904) with different m/v concentrations (20%, 10%, 4%) in phenol-red free medium 777 were freshly prepared. Right before imaging, the normal cell culture medium was 778 changed into 1 ml of phenol-red-free medium in the 35 mm dish. After starting the image 779 acquisition, we added 1 ml of pre-warmed hexanediol stock solution (20%, 10%, 4%) to 780 the 35 mm dish without pausing imaging to adjust to the final concentrations of 10%, 5%, 781 or 2%. We treated the time of hexanediol addition as the time "0" when quantification of 782 surviving puncta in Fig. 1E. 783 784 Acknowledgments 785 We thank Dr. Anthony A. Hyman for sharing the plasmids; thank Jie Wang for advice 786 and sharing the scripts for statistical analysis of partition coefficient and the amount 787 . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 6, 2021. was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 6, 2021.      was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 6, 2021.   was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.06.451310 doi: bioRxiv preprint . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 6, 2021. was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.06.451310 doi: bioRxiv preprint 1340 1341 was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 6, 2021. was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.06.451310 doi: bioRxiv preprint 1353 1354 was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.06.451310 doi: bioRxiv preprint . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.06.451310 doi: bioRxiv preprint Figure 4. IDR-driven YBX1 phase separation is required for sorting YBX1 into exosomes.

1391
. CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.06.451310 doi: bioRxiv preprint . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.06.451310 doi: bioRxiv preprint . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.06.451310 doi: bioRxiv preprint 1433 1434

1436
. CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.06.451310 doi: bioRxiv preprint . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.06.451310 doi: bioRxiv preprint . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 6, 2021. was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 6, 2021. was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 6, 2021. ; https://doi.org/10.1101/2021.07.06.451310 doi: bioRxiv preprint