Genome-wide interrogation of extracellular vesicle biology using barcoded miRNAs

Extracellular vesicles mediate transfer of biologically active molecules between neighboring or distant cells, and these vesicles may play important roles in normal physiology and the pathogenesis of multiple disease states including cancer. However, the underlying molecular mechanisms of their biogenesis and release remain unknown. We designed artificially barcoded, exosomal microRNAs (bEXOmiRs) to monitor extracellular vesicle release quantitatively using deep sequencing. We then expressed distinct pairs of CRISPR guide RNAs and bEXOmiRs, enabling identification of genes influencing bEXOmiR secretion from Cas9-edited cells. This approach uncovered genes with unrecognized roles in multivesicular endosome exocytosis, including critical roles for Wnt signaling in extracellular vesicle release regulation. Coupling bEXOmiR reporter analysis with CRISPR-Cas9 screening provides a powerful and unbiased means to study extracellular vesicle biology and for the first time, to associate a nucleic acid tag with individual membrane vesicles.

); this was also true when comparing two independent or replicate screens as 95 described below (see below Figure 3A, B).

97
The overall representation of bEXOmiRs in EV fractions suggested that intrinsic 98 differences in barcode sequences to some extent, influenced exosomal targeting, as 99 seen for other miRNAs (5, 10, 11). Some bEXOmiRs were very efficiently detected in 100 EVs while others were not ( Figure 1D). This does not reflect differences in synthesis,  A CRISPR/Cas9 screen using bEXOmiRs 115 We next used bEXOmiRs in a pooled, genome-wide CRISPR/Cas9-mediated screening 116 protocol to monitor EV release ( Figure 2A). Briefly, unique CRISPR single guide 117 (sg)RNAs linked by synthesis with distinct bEXOmiRs are co-expressed; we then 118 determine how knockout of every gene influences EV release by comparing bEXOmiR 119 barcode abundance in EV fractions isolated from wild type compared with Cas9-edited 120 cells ( Figure 2A). Our hope was that the use of 10 CRISPR sgRNAs per gene, coupled   Altogether, this analysis (9 duplicated sub-library screens) required >100 liters of EV 154 preparations and associated, large-scale miRNA sequencing runs. Figure 3A,B shows 155 a comparison of total bEXOmiR abundance in EV analyses from two replicate screens 156 that similar to our test bEXOmiR ( Figure 1D), also showed excellent concordance 157 (R 2 =0.95). Shown in Figure 3C is a comparison of total bEXOmiR abundance in cells 158 versus EVs. There was essentially no correlation between these pools (R 2 =0.13-0.17), 159 consistent with differential sorting of bEXOmiRs into EVs, as observed for endogenous

240
In an attempt to determine more precisely, the pathway(s) impacted by several of the 241 key hits identified, we made use of a TIRF microscopy-based live cell assay to monitor         Bafilomycin A1 treatment, that activates EV release, actually inhibits Wnt signaling (47).

344
It is very satisfying that the very EV pathway that may generate functional Wnt signals 345 (34) is in fact directly regulated by Wnt signaling via an important and yet to be 346 characterized feedback mechanism. Altogether, complex crosstalk between signaling 347 pathways may influence EV secretion in ways that will be important to define in future 348 work.   supplemented with 20% FBS at 100,000 X g in a 45Ti rotor, as described (49).  EV isolation--K562 suspension cell cultures were seeded at 5 × 10 5 cells/ml in EV-free 430 RPMI and grown for 24-48 h. Cell cultures were centrifuged at 300 × g, 5 min to pellet 431 cells, and supernatants were centrifuged again at 2100 × g, 20 min to pellet dead cells.

432
The resulting supernatant was spun again at 10,000 × g, 30 min to pellet cell debris and representation in WT exosomes as to interfere with sequencing depth (count > 10,000).

491
Additional barcodes were randomly generated as above for larger libraries.     edges whose width is proportional to a combined significance score (see Methods).

989
Activators are shown in red, suppressors in blue. Node color intensity is proportional to 990 effect size found in the screen. Node size is proportional to significance score (-Log p-991 value) calculated in the screen analysis. In all panels, asterisks indicate that an isoform 992 or close family member has been shown to play a role in the same category.