Validation of the purification procedures

A. Workflow for TNT vs. EVP purification. EVPs were purified from cell culture supernatant, TNTs from remaining attached cell supernatant after shaking. * indicates the fraction used for nanoFCM analysis.

B. Representative pictures of negative staining and Transmission Electron Microscopy from EVP and TNT fractions as indicated. Scale bars are 100, 500 and 200 nm respectively.

C. Violin plot of the size distribution of EVP diameters (124 vesicles) and TNT diameters and lengths (40 objects), line is the median. EVP and TNT diameter means are 60 and 69 nm respectively, TNT lengths extend from 140 to 912 nm, mean at 372 nm. Statistical analysis is one-way Anova with Tukey post-hoc correction. Ns, non-significant, ****p<0.0001.

D. Schematic representation of stable cell lines where green color indicates the location of GFP-tagged protein: H2B-GFP (nuclear), actin chromobody-GFP (actin cytoskeleton including TNTs) and GFP-CD9 (cell surface, TNTs, EVPs).

E. Scatter dot plot representing the mean percentage (with SD) of GFP-positive particles analyzed by nanoFCM. Statistical analysis of 3 independent experiments to 6 for GFP-CD9 (oneway Anova with Tukey post-hoc correction) show the following respective p-values (from top): blue*: 0.0252, red**: 0.0089, red*: 0.0131, black*: 0.0161. Means values are (from left to right): 0.01, 0.3, 4.8, 1.75, 0.16, 37.07, 27.7 and 11.6.

F. Western blot of WCE (20μg, corresponding to around 0.1×106 cells), TNT and EVP fractions (both from 10 106 cells) prepared from the same cells, blotted with CD63, CD9, α-tubulin and GM130 specific antibodies. White lane indicates that intervening lanes of the same gel (and same exposure) have been spliced out.

Analysis of the TNTome

A. Proteomap of the 1177 proteins of the TNTome, sorted in 4 quartiles depending on their mean iBAQ. Protein accession and mean iBAQ were used to create ProteoMap, analyzed according to Gene Ontology.

B. STRING physical association network for the cytoskeleton-related proteins listed in table S2. Color groups were created using Cytoscape. Green are microtubule-related proteins, blue are intermediate filaments, orange are actin-interacting proteins. Blue and pink edges show physical interactions based on databases and experiments respectively.

C. Full STRING functional association network for integral surface membrane proteins of TNTome, based on table S3. Orange are Integrin proteins, red are Ephrin receptors, dark green are Cadherins, light green are Sodium/potassium transporting ATPase ions channels, purple are monocarboxylate and amino acids transporters, and yellow are tetraspanin-related proteins.

D. Volcano plot of the mass spectrometry analysis based on the 4 EVP and TNT preparations, showing the maximum log2(Fold-change) in x-axis measured between TNT and EVP fractions and the corresponding -log10 (p-value) in y-axis. Dashed lines indicate differential analysis quadrants with log2 (Fold-change) =0.58 and false discovery rate FDR = 1%. Common EVP=TNT are non-significantly different (FDR >0.05) with FC>1.5, and FC<1.5. Each quadrant is named above and the number of identified proteins is indicated. Left and right are proteins non overlapping in both fractions: TNT-only and EVP only. Note that in EVP-only fraction, 10 proteins were found in TNTome (based on 12 experiments) and should therefore be removed. For the TNT proteins, only the proteins also present in TNTome have been counted.

Expression of CD9/CD81 in TNTs and effects of their overexpression or invalidation

A. Immunofluorescence of CD9 (green) and CD81 (red) in SH-SY5Y cells. Cells were also stained with phalloidin (magenta) and DAPI (blue) to visualize actin and nuclei. This representative image corresponds to the 4th to 5th slices of a stack comprising 11 slices (1 being at the bottom). In A, B and D, yellow arrowheads point to TNTs that are visible in the shown pictures, however more TNTs were counted over the whole stack, which were not annotated on the figure. Scale bars correspond to 10 µm.

B. Representative images of TNT-connected cells in WT, CD9 KO, CD81 KO and CD9&CD81 KO cells (max projection of 4 to 7 upper slices, z-step 0.4 µm), stained with WGA-488 (green) to label the membrane and DAPI (blue) to label the nuclei.

C. Graph of the % of TNT-connected cells in the different cells, from 3 independent experiments. Mean and standard deviation (SD) are: WT = 31.6 ± 1.25; CD9 KO = 25.2 ± 1.24; CD81 KO = 30.4 ± 1.65; CD9 & CD81 KO = 16.3 ± 4.81. p-values are indicated above the brackets in all graphs.

D. Representative images (max projection of 5 to 7 upper slices, z-step 0.4 μm) of TNT-connected cells in WT, CD9 OE and CD81 OE cells, stained as in B.

E. Graph of the % of TNT-connected cells in the indicated cells from 3 independent experiments. Mean ± SD are: WT = 29.8 ± 1.11; CD9 OE = 45.3 ± 3.17; CD81 OE = 29.5 ± 0.84.

F. Vesicle transfer assay from donor cells KO of CD9 and CD81, as schematized on the left. Graphs are % of acceptor cells containing DiD vesicles from 3 independent experiments. Mean ± SD are: WT = 32.7 ± 10.25; CD9 KO = 18.4 ± 8; CD81 KO = 19.1 ± 4.86; CD9 & CD81 KO = 10.5 ± 1.52.

G. Vesicle transfer assay from donor cells OE CD9 or CD81, as schematized on the left. Graphs are % of acceptor cells containing DiD vesicles from 3 independent experiments. Mean ± SD are: WT = 29.3 ± 7.45; CD9 OE = 61.3 ± 20.44; CD81 OE = 55.6 ± 10.31.

CD9 and CD81 act successively in the formation of TNTs

A. Representative images of WT, CD9 OE, CD9 OE + CD81 KO, CD81 OE and CD81 OE + CD9 KO cells, stained with WGA-488 (green) and DAPI (blue). Yellow arrowheads show TNTs. Scale bars correspond to 10 µm.

B. Graph of the % of TNT-connected cells in the tetraspanin OE + KO cells. Mean ± SD (N=3) are: WT = 28.6 ± 2.05; CD9 OE = 44.2 ± 2.4; CD9 OE + CD81 KO = 48.2 ± 2.16; CD81 OE = 29.5 ± 1.36; CD81 OE + CD9 KO = 20.2 ± 2.09. p-values are above the brackets.

C. Coculture between tetraspanin OE + KO cells used as donors and WT GFP cells used as acceptors, as schematized on the left. The graph shows the % of acceptor cells containing DiD vesicles. Mean ± SD from 3 independent experiments: WT = 26.1 ± 8.94; CD9 OE = 73.8 ± 12.64; CD9 OE + CD81 KO = 87.7 ± 5.11; CD81 OE = 73 ± 14.60; CD81 OE + CD9 KO = 45.4 ± 11.53.

CD9 and CD46 antibody treatment in WT and CD81 KO cells

A. Representative confocal images of Actin chromobody-GFP expressing WT cells, treated for 3 hours with either anti-CD9 antibody coupled to Alexa Fluor 568 (upper row), or with anti-CD46 antibody coupled to to Alexa Fluor 594 (second row). Alexa-647 coupled WGA and DAPI were added post-fixation. The images correspond to a maximal projection of 3 or 4 slices of the z-stack encompassing the TNTs indicated with yellow arrowheads. The scale bars correspond to 10 µm.

B. Graph of the % of TNT-connected cells in NT, CD46 AB or CD9 AB treatment in WT cells. Mean percentages ± SD (N=3) are 30.4, 31.9 and 41.8, p values resulting from statistical analysis are indicated above the brackets.

C. Top, schematic of the antibody treatment experiment after coculture of WT SH-SY5Y donor cells (DiD in yellow circles to stain the vesicles) and WT GFP-labeled acceptor cells, treated with control antibodies (CTR AB) or with antibodies anti-CD9 (CD9 AB) for an additional 2 hours. Graph represents the % of acceptor cells containing DiD vesicles of the cocultures of WT cells with CTR AB or CD9 AB treatment. Mean ± SD (N=3) are: CTR AB = 27.7 ± 2.48; CD9 AB = 36.8 ± 3.03.

D. Representative confocal images as in A except that CD81 KO cells were used.

E. Graph of the % of TNT-connected cells in NT, CD46 AB or CD9 AB treatment in CD81 KO cells. Mean percentages ± SD (N=3) are 29.3, 28.5 and 41.4, p values resulting from statistical analysis are indicated above the brackets.

F. Top, schematic of the antibody treatment experiment after coculture of CD81 KO SH-SY5Y donor cells (DiD in yellow circles to stain the vesicles) and WT GFP-labeled acceptor cells, treated with control antibodies (CTR AB) or with antibodies anti-CD9 (CD9 AB) for an additional 2 hours. Graph represents the % of acceptor cells containing DiD vesicles, mean percentages ± SD (N=3) are: CTR AB = 20.6 ± 1.67; CD9 AB = 20.6 ± 2.16

Stabilization of TNTs by CD9 AB.

A. Representative snapshots of a TNT over time from Movie S1, corresponding to non-treated WT SH-SY5Y cells (WT NT). In A-F, green panels correspond to the signal of actin chromobody-GFP (AC GFP). Yellow arrowheads in the green panels are pointing to TNTs and the yellow crosses mark the breakage/dissociation of the TNT. Yellow stars in red panels mark the accumulation of CD9 AB (at the base/tip of the TNTs) or CD46 AB (intracellularly). Scale bars correspond to 10 µm.

B. Representative snapshots of TNTs over time from Movie S2, corresponding to WT SH-SY5Y cells treated with 10 µg/mL of CD9 specific antibody (WT + CD9 AB). Red panel corresponds to the signal of CD9 antibody coupled to Alexa Fluor 568.

C. Representative snapshots of TNTs over time from Movie S3, corresponding to WT SH-SY5Y cells treated with 10 µg/mL of CD46 specific antibody (WT + CD46 AB), coupled to Alexa Fluor 594 (red panel).

D. Representative snapshots of a TNT over time from Movie S4, corresponding to non-treated CD81 KO SH-SY5Y cells (CD81 KO NT).

E. Representative snapshots of TNTs over time from Movie S5, corresponding to CD81 KO SH-SY5Y cells treated with 10 µg/mL of CD9 specific antibody (CD81 KO + CD9 AB), coupled to Alexa Fluor 568.

F. Representative snapshots of TNTs over time from Movie S6, corresponding to CD81 KO SH-SY5Y cells treated with 10 µg/mL of CD46 specific antibody (CD81 KO + CD46 AB) coupled to Alexa Fluor 594.

G. Average duration of TNTs in WT (left) or CD81 KO cells (right), measured by live imaging from 3 independent experiments, and represented in Violin plots (with line at median). Left: mean lifetime of 33 TNTs in WT non-treated (NT) cells was 10.5 minutes (± 5.59), 12.8 minutes (± 11.68) in 27 TNTs measured for WT cells treated with CD46 AB, while for WT cells treated with CD9 AB the average duration of the 40 TNTs measured was 22.6 minutes (± 10.87). Right: mean lifetime of TNTs in CD81 KO non-treated (NT) cells, CD81 KO cells treated with CD46 AB, and CD81 KO cells treated with CD9 AB. The average lifetime of 31 TNTs in CD81 KO NT cells was 11.7 minutes (± 7.78), 15.3 minutes (± 11.70) in 27 TNTs in CD81 KO cells treated with CD46 AB, while the average duration of 33 TNTs in CD81 KO cells treated with CD9 AB was also 28.2 minutes (± 15.66). Statistical analysis was performed using one-way Anova with Holm-Sidak’s multicomparison test.

Completion of TNTs by CD81

A. DiD vesicles in TNTs from WT or CD81 KO cells. Actin chromobody-expressing SH-SY5Y cells, either WT or CD81 KO, were challenged with DiD for 30 min and fixed to preserve TNTs after 8 or 24 hours. After additional WGA staining, TNTs (above 10 µm in length) were identified by confocal microscopy using WGA channel, and classified according to their content in DiD vesicles. Left is shown representative examples of long TNTs containing (projection of 12 slices of 0.19 µm) or not (projection of 7 slices) DiD-positive vesicles, from CD81 KO and WT cells respectively. Scale bars are 10 µm, yellow arrowheads point to DiD-positive vesicles in the TNT. The floating bar graph on the right shows the percentage of TNTs that contain DiD-positive vesicles from 6 independent experiments (min to max results, line at means (30.3 and 53.0), medians are 33.4 and 53.7. Statistical analysis is Mann-Whitney test, p-value is above the brackets. TNT number per experiment: 21, 23, 19 (8h of treatment), 6, 30, 41 (24h of treatment) for WT cells; 22, 17, 29 (8h of treatment), 17, 22, 29 (24h of treatment) for CD81KO cells. See also movies S7, S8, S9.

B. Coculture of actin chromobody-GFP-expressing cells (WT or CD81 KO) with WT cells. After fixation, cells were additionally labeled with WGA (Alexa 647, in white) and Phalloidin-Rhodamin (red). TNTs were identified by confocal imaging and classified as schematized above the diagrams: fully green when red and green signals overlapped throughout the structure (exemplified in the first row of pictures), partially or not green when green signal was interrupted or not present at all in the TNTs (exemplified in second and third rows respectively). First and third rows are maximal projections of 6 and 7 slices (of 0.19 µm) encompassing the TNT of WT-WT coculture, second row is a maximal projection of 5 slices of CD81 KO-WT coculture. Yellow arrowheads point to TNTs. Scale bars are 10 µm.

C. Quantification of the percentage of fully green TNTs in the cocultures of panel B, from 5 independent experiments (approximately 20 TNTs per condition and experiment, in total 91 and 92 analyzed TNTs from WT-WT and CD81 KO-WT cocultures respectively). Floating bars are min to max results, line at means (69.3 and 34%), medians are 66.7 and 27.8 %, statistical analysis was Mann-Whitney test.

D. Working model of CD9/CD81 roles on TNT formation. TNTs (black dotted-line frame) are made up of iTNTs, supported by actin (red arrows). In WT cells, CD9 and CD81 are present on growing iTNT (probably together with additional interactors, indicated as yellow circle), and CD9 stabilizes them. At the junction between iTNT and opposing cell membrane, CD81 participates in membrane docking/ fusion, which results in the full completion of an open, functional TNT (right panel). When cells are treated with CD9AB, the active complexes are further enriched/stabilized, and more TNTs are stabilized. In CD9 KO cells, growing iTNTs are not stabilized, resulting in the decrease of TNT number. In CD81 KO cells, CD9-containing complexes (green circles) are still active for stabilization of iTNTs but anchoring of the iTNTs to opposite cell/fusion is no longer induced by CD81, resulting in apparent, but non-functional TNTs, in which vesicles (purple circles) are trapped.