Exosome markers localize to the base and tips of filopodia in cancer cells and cortical neurons.

A. Representative confocal image of HT1080 cells stained with phalloidin-Alexa fluor 488 and CD63 shown in red. The red channel has been edited using brightness and contrast tools for ease of visibility. Note localization of the exosome marker CD63 in extracellular deposits and at or near the tips of filopodia (arrowheads). Representative of 20 images. Scale bar is 10 μm in each panel. B. Time series of pHluorin-M153R-CD63-mScarlet movie in HT1080 cells. Yellow arrowheads indicate fusion sites and yellow arrows indicate filopodia. Note a filopodium forming shortly after MVE fusion. C. Representative kymographs showing MVE docking (red), fusion (yellow), and filopodia formation in HT1080 cells. Yellow arrowheads denote MVE fusion events and black arrowheads denote the formation of a filopodium. Each pixel is 10 seconds x 0.2857 μm. D. Quantification of the time elapsed between MVE fusion and filopodia formation. n=420 kymographs from 46 cells from 3 independent experiments. E. Primary cortical neurons were co-transfected with GFP-Rab27b (green) and mCherry as a filler to visualize filopodia (red) on DIV5 and fixed for imaging on DIV6. SV2 negative staining (no signal) identifies these structures as filopodia instead of dendritic spines. Arrows in merged images indicate localization of GFP-Rab27b to tips and bases of filopodia. Scale bars = 5 µm. F. Percent GFP-Rab27b localization to tips and bases of filopodia in 70 individual cortical neurons from three independent experiments. Red line indicates the median. Error bars, SEM. ns, not significant; * p<0.05; ** p<0.01; *** p<0.001

Exosomes promote filopodia formation and stability in cancer cells.

A. EVs secreted from equal numbers of control (shLacZ), Rab27a-KD, and Hrs-KD B16F1 cells over 48 h were quantified using NanoSight particle tracking analysis (N=3). B. Representative images of control (shLacZ), Hrs-KD, and Rab27a-KD B16F1 cells stained with rhodamine-phalloidin. Arrowheads show examples of filopodia. Images have been edited with brightness and contrast for ease of visibility. C. Quantification of filopodia from images as in B (N=3, ≥ 27 total cells per condition). Filopodia number per 500 μm2 cell area. D. Representative images of filopodia in B16F1 control (shLacZ) and exosome-depleted (shHrs) cells treated for 18 h with LEVs or SEVs isolated from control cells. Arrowheads show examples of filopodia. Images have been edited with brightness and contrast for ease of visibility. E. Quantification of filopodia from images as in D (N=3, ≥ 20 total cells per condition). F. Filopodia number in B16F1 shScr cells treated with indicated numbers of purified LEVs or SEVs, for 18 hours (N=3, ≥25 total cells per condition). G. Control (shLacZ) and exosome-depleted (shHrs) B16F1 cells were transiently transfected with tdTomato-F-tractin to visualize filopodia formation. Live images were taken every 30 seconds for 15 minutes and newly formed filopodia were counted at each time point. Only filopodia that form and fully retract during the duration of the video were quantified. (N=3, β20 total cells per type) H. Lifetime of newly formed filopodia from (G). Lifetime is defined as the time from first formation of the filopodia to full retraction. Bars represent mean and error bars are SEM. (N=3, β20 total cells per type) Scale bars in wide field and zoom insets = 10 μm. Error bars, SEM. ns, not significant; * p<0.05; ** p<0.01; *** p<0.001.

Exosomes promote filopodia, spine, and synapse formation in cortical neurons.

Primary cortical neurons were co-transfected with plasmids for expression or inhibition of exosome regulatory molecules and mCherry (red) as a cytoplasmic filler to image neuronal protrusions, then fixed and immunostained with SV2 (pseudo-colored in cyan) to visualize synapses. Filopodia were identified as thin SV2-negative, mCherry-positive protrusions. Spines were identified as dendritic protrusions that co-localize with SV2. Synapses were identified as SV2-positive puncta present on both dendritic protrusions and dendritic shafts. A, C. Representative images of primary rat cortical neurons co-transfected at DIV6 when filopodia typically form (A) or DIV12 when synapses typically form (C) with GFP or GFP-Rab27b (green, left images) and mCherry (red) and co-stained with SV2 (blue, right images). B, D. Quantification of filopodia (DIV6), spine and synapse density (DIV12) from images as in A and C. E-G. Images and analysis from neurons transfected with control shRNA (NTshRNA) or shRNAs against Rab27b (E) or Hrs (G) and immunostained for SV2. Quantification of filopodia (DIV6), spine and synapse density (DIV12) for Rab27b-KD (F) or Hrs-KD (H) neurons. I, J. Rescue experiments. Control and KD neurons (as indicated) expressing shRNAs and mCherry were treated with purified SEVs on DIV5 for 24 hrs at a dose of 200 EVs per neuron, then fixed and stained for SV2 at DIV6. Filopodia, spine and synapse density were quantified from more than 30 primary or secondary dendritic shafts from three independent experiments for each condition. Scale bars = 5 μm. Error bars, SEM. *p<0.05, **p<0.01, ***p<0.001.

Endoglin is an SEV-enriched cargo that promotes filopodia formation.

A. Purified LEVs and SEVs were run on a colloidal blue-stained gel. Four arrows denote SEV bands that were cut and submitted for proteomics, along with notable proteins identified (see Supplemental Table 1 for the full proteomics results). B. B16F1 total cell lysate (TCL), LEVs, and density gradient purified SEVs were run on an SDS-PAGE gel and probed by Western blot for endoglin, and EV positive (HSP70, TSG101, flotillin-1, and CD63) and negative (GM130) markers. C. Total cell lysate (TCL) and small EVs (SEVs) from endoglin-KD (shEng) and control (shScr) B16F1 cells were run on an SDS-PAGE gel and probed by Western blot for endoglin, EV marker TSG101, and EV negative marker GM130. D. Representative images from control (shScr) or endoglin-KD (shEng) B16F1 cell lines incubated for 18 h with no EVs (left panels), or with SEVs purified from control (+shScr SEVs) or shEng cell lines (+shEng SEVs) (right panels). Arrowheads indicate example filopodia. Scale bar = 10 μm. E. Quantification of filopodia in control (shScr) and endoglin knockdown (shEng) cells treated with the indicated number of LEVs or SEVs for 18 hours (N=3, at least 20 cells per condition per repetition). F. Filopodia number in B16F1 control (shLacZ) or exosome depleted (shHrs) cells treated with indicated numbers of LEVs, control (shScr) SEVs, or endoglin-KD (shEng1) SEVs for 18 hours. N=3, ≥ 20 cells per condition per rep. Representative images for this experiment are shown in figure S5E. G and H. B16F1 cells were transfected with tdTomato-F-Tractin and imaged live every 30 seconds for 15 minutes. Only filopodia that form and fully retract during the duration of each video were quantified. G. De novo filopodia formation. H. Filopodia lifetime, defined as the time from initial filopodia formation to full retraction. Bars represent mean and error bars are SEM. (N=3, β25 total cells per type) ns, not significant; * p<0.05; ** p<0.01; *** p<0.001.

Exosomal endoglin controls motility and metastatic colony formation.

A. Cartoon diagram of metastatic colony assay in avian embryos. On day 0, fluorescent HT1080 cells were injected (100,000 cells per egg) into the vein of the chicken embryo. On day 4, the egg was opened, the embryo was sacrificed, and a circular tool was used to punch holes through the shell. The chorioallantoic membrane (CAM) was peeled away from the shell, placed on a glass slide with a coverslip, and immediately imaged. The cartoon was created using BioRender.com/o3qeu33. B. Representative low power wide field images of colony formation in the CAM. Scale bar = 200 μm. C. Representative high power wide field images of colony formation in the CAM. Scale bar = 100 μm. D and E. Quantification of CAM colony number (D) and size (E) from high power images as in C. 4-7 eggs were harvested per condition. D. Colony number is graphed per field of view using 25-30 fields of view per egg. E. Quantification of the percent of large (≥ 5000 μm2) colonies formed by control and shEng HT1080 cells. F. 3D invasion in collagen. HT1080 cell spheroids were seeded in collagen gels and imaged for 8 hours. Invasion is quantified as fold area increase in the size of each spheroid over 8 hours. Scale bar = 100 μm. Error bars, SEM. ns, not significant; *p<0.05; ** p<0.01; *** p<0.001.

Exosomal endoglin promotes filopodia formation through THSD7A.

A. Native gel Western blot of B16F1 SEVs. B. Standard Western blot of HT1080 SEVs. C. Western blot of cortical neuron total cell lysate (TCL) and SEVs. D. Representative images and quantitation of filopodia number in control (lipofectamine) and THSD7A-mScarlet-transfected HT1080 cells. Arrowheads indicate THSD7A at the ends of filopodia (white arrowheads) or in extracellular deposits (red arrowheads). Scale bars in wide field and zoom insets = 10 μm. E. (left) Western blot of control shRNA (NTC) and shTHSD7A (C-04, C05, C-06) - expressing HT1080 cell lines. Vinculin is used as a loading control and numbers below the blot indicate normalized THSD7A levels. (right) Filopodia counts in control and shTHSD7A HT1080 cells. N=3, at least 20 cells per condition per rep. F. THSD7A coated coverslips rescue filopodia defect in shEng B16F1 and HT1080 cells. N=3, at least 20 cells per condition per rep. G. and H. Cortical neurons were transfected with a FLAG-THSD7A expression vector 77 or vector control, fixed and stained with an antibody against THSD7A, and imaged by confocal microscopy. G. Representative images. Arrows indicate THSD7A localization to the tips of filopodia. Scale bar = 5 μm. H. Quantification of filopodia in neurons expressing FLAG-THSD7A or control vector. n=42 neurons from three separate experiments. I. Rescue of filopodia numbers in shHrs neurons plated on dishes coated with various concentrations of recombinant human THSD7A, as indicated. Error bars, SEM. ns, not significant; * p<0.05; ** p<0.01; *** p<0.001.

Endoglin controls THSD7A trafficking to exosomes.

A. Western blot analysis of total cell lysates (TCL) and SEVs from HT1080 control and shEng cells +/-rescue with WT endoglin or control expression vectors. The figure was made from cropped images of membranes to remove irrelevant lanes. B. Quantification of endoglin expression (normalized to flotillin-1 as a loading control, and relative to shScr control) from triplicate Western blots as in (A). C. Quantification of THSD7A expression (relative to flotillin-1 as a loading control, and relative to shScr control) from triplicate Western blots as in (A). D. Quantification of filopodia in HT1080 control cells and shEng cells rescued with WT endoglin expression. N=3, at least 30 total cells per condition. E. Representative confocal images of THSD7A-mScarlet-expressing control and shEng HT1080 cells immunostained for CD63. Box 1 shows extracellular THSD7A and CD63 deposits. Box 2 shows intracellular CD63-positive MVEs. For both boxes, the zoomed images have been adjusted for brightness and contrast (to equivalent levels for control and shEng cells) for easy visualization. Note that overlap of THSD7A (magenta) and CD63 (green) gives a white signal, pointed out with white arrowheads in shEng merged image in Zoom 2. Scale bar is 10 μm in wider field view and 5 μm in zoom insets. F. Quantification of colocalization of internal CD63 and mScarlet signals in HT1080 cells from nonadjusted images. N=3, ≥ 20 cells per condition per rep. Error bars, SEM. ns, not significant; * p<0.05; ** p<0.01; *** p<0.001.

Filopodia formation induced by THSD7A depends on Cdc42 activity.

Control and endoglin-KD HT1080 cells were plated on coverslips coated with poly-D-lysine (PDL) or THSD7A. In some cases, cells were treated with the Cdc42 inhibitor ML141 (10 µM) or transfected with the dominant active Cdc42 mutant Q61L, as indicated. N=3, ≥ 20 cells per condition per rep. Error bars, SEM. ns, not significant; * p<0.05; ** p<0.01; *** p<0.001.

Model of exosome-induced filopodia formation.

A. In tumor cells, endoglin and THSD7A are trafficked into intralumenal vesicles (ILV) in multivesicular endosomes (MVEs) for secretion. Inhibiting the exosome biogenesis pathway by blocking Hrs or inhibiting MVE docking by blocking Rab27a subsequently reduces exosome secretion and filopodia formation. SEVs carrying THSD7A can induce filopodia on target cells via Cdc42, leading to increased cell motility and metastasis. When endoglin levels are lowered (such as by KD), THSD7A is retained inside cells in CD63-positive endosomes and its levels in SEVs are greatly decreased. The drop in THSD7A levels in endoglin-KD EVs could be due either to a lack of trafficking into ILVs or, alternatively, enhanced lysosomal degradation of THSD7A-containing MVEs. Given that THSD7A accumulates in CD63-positive endolysosomal compartments in endoglin-KD cells (Fig 7E,F), the latter possibility seems more likely. The cartoon was created using BioRender.com/r10h575. B. In primary neurons, exosome biogenesis is controlled by the formation of ILVs by Hrs and MVE docking is controlled by Rab27b. Knockdown of either of these proteins results in reduced formation of filopodia, dendritic spines, and synapses in both cortical and hippocampal neurons. Similar to cancer cells, THSD7A is carried in neuronal SEVs and induces filopodia. The cartoon was created using BioRender.com/7ljyfta.