PTEN deficiency exposes a requirement for an ARF GTPase module for integrin-dependent invasion in ovarian cancer

Abstract

Dysregulation of the PI3K/AKT pathway is a common occurrence in high-grade serous ovarian carcinoma (HGSOC), with the loss of the tumour suppressor PTEN in HGSOC being associated with poor prognosis. The cellular mechanisms of how PTEN loss contributes to HGSOC are largely unknown. We here utilise time-lapse imaging of HGSOC spheroids coupled to a machine learning approach to classify the phenotype of PTEN loss. PTEN deficiency induces PI(3,4,5)P₃ -rich and -dependent membrane protrusions into the extracellular matrix (ECM), resulting in a collective invasion phenotype. We identify the small GTPase ARF6 as a crucial vulnerability of HGSOC cells upon PTEN loss. Through a functional proteomic CRISPR screen of ARF6 interactors, we identify the ARF GTPase-activating protein (GAP) AGAP1 and the ECM receptor β1-integrin (ITGB1) as key ARF6 interactors in HGSOC regulating PTEN loss-associated invasion. ARF6 functions to promote invasion by controlling the recycling of internalised, active β1-integrin to maintain invasive activity into the ECM. The expression of the CYTH2-ARF6-AGAP1 complex in HGSOC patients is inversely associated with outcome, allowing the identification of patient groups with improved versus poor outcome. ARF6 may represent a therapeutic vulnerability in PTEN-depleted HGSOC.

Keywords: 3D spheroids; ARF6; Ovarian Cancer; PTEN; integrins.

Figure 1. Loss of Pten in HGSOC epithelium is associated with poor outcome

1. A–C

   PTEN mRNA levels in LCM normal ovarian surface epithelium versus high‐grade serous ovarian cancer (HGSOC) epithelium or normal ovarian stroma versus ovarian cancer‐associated stroma. Data sets; (A) GSE40595, (B) GSE38666, (C) Epithelium only, GSE14407. Sample size (n) and P‐values, (Mann–Whitney) annotated, whiskers Min–Max, line at median.

2. D

   PTEN mRNA levels in normal ovarian surface epithelium versus tumour. Data set ID, sample size (n) and P‐values (Mann–Whitney) annotated, whiskers Min–Max, line at median.

3. E

   Copy Number, mRNA, protein level changes and mutations identified across PTEN and TP53 in the TCGA data set of OC. Sample size (n) = 414 patients.

4. F

   Overall survival (% patients, months) of OC patients. Highest quartile (Q4) versus combination of quartiles 1–3 (Q1 + 2 + 3), PTEN mRNA (TCGA, OV). Median survival (40 and 44 months), sample size (n) and P‐value, Log‐rank test (Mantel–Cox) annotated.

5. G

   Overall survival (% patients, months) of OC patients. Highest quartile (Q4) versus combination of quartiles 1–3 (Q1 + 2 + 3), PTEN protein. Reverse Phase Protein Array Data, TCGA OV. Median survival (46 and 57 months), sample size (n) and P‐value, Log‐rank test (Mantel‐Cox).

6. H

   Differential abundance (x, Log Ratio between conditions; y, −Log₁₀ q‐values) of Reverse Phase Protein Array data (TCGA, OV) in patient grouped by PTEN mRNA, High (Q4) versus Low (Q1). Significant, blue (−Log₁₀ q‐values > 1.3); AKT signalling pathway, labelled.

7. I

   Differential abundance (x, Log Ratio between conditions; y, −Log₁₀ q‐values) of proteins in PTEN High (Q4) versus PTEN Low (Q1) protein samples. Reverse Phase Protein Array Data, TCGA OV. Significantly altered components in AKT signalling pathway labelled (−Log₁₀ q‐value > 1.3).

Source data are available online for this figure.

Figure EV1. Characterisation of Pten loss effect on PI3K‐AKT in 2D culture

1. A

   Schema, derivation of Pten and Trp53 alterations in ID8 sublines.

2. B, C

   Western blot in ID8 sublines. (B) TRP53, PTEN, GAPDH expression upon Nutlin‐3A (MDM2 inhibitor) treatment to stabilise P53 or (C) pS473‐AKT, AKT, Vinculin (VCL) expression. Each panel is representative of n = 3 lysate preparations for each subline. GAPDH and VCL are loading controls for each panel.

3. D

   Quantitation of (C). Data, mean ± SD of pS473‐AKT/total AKT intensity ratio, normalised to WT. Unpaired, two‐tailed t‐test; P‐values, annotated.

4. E

   Western blot, TRP53, PTEN, VCL in ID8 Wild Type cells expressing non‐targeting (sgNT) or Pten‐targeting sgRNA upon Nutlin‐3A (MDM2 inhibitor) treatment. Representative of n = 3 lysate preparations for each subline. VCL is loading control.

5. F

   Quantitation of PTEN band intensity from (E). Data, mean ± SD of band intensity, normalised to ID8 Wild‐Type sgNT. Unpaired, two‐tailed t‐test; P‐values, annotated.

6. G

   Western blot, pAKT(S473), AKT pan, VCL in ID8 Wild Type cells expressing non‐targeting (sgNT) or Pten‐targeting sgRNA. Representative of n = 3 lysate preparations for each subline. VCL is loading control.

7. H

   Quantitation of (G). p:t AKT ratio. Data, mean ± SD of band intensity, normalised to ID8 Wild‐Type sgNT. Unpaired, two‐tailed t‐test; P‐values, annotated.

8. I

   Quantitation of (J). Data, ratio pS473‐AKT signal at indicated regions to total area. Means, overlaid on plots of all data points (exact number per replicate provided in Table EV1) as distinctly coloured dots according to culture replicate number. P‐values are annotated, ANOVA with Tukey's honest significant difference (HSD) test.

9. J

   ID8 cells plated in 2D, stained with pS473‐AKT (grey) and Hoechst (Magenta) (bottom panels), segmented into indicated regions (perinuclear, cytoplasmic, membrane) (top panels). Colour in selected cells panel: red, excluded due to touching image edge, green, included for segmentation. Arrowheads, pS473‐AKT at cell membrane. Scale bar, 20 μm. N = 3 independent experiments, four technical replicates/subline/experiment. Total cell number per condition, Table EV1.

10. K

    Percentage of cells classified as Round (green), Cobblestone (red) or Elongated (blue) in Wild Type ID8 cells. Classification of ID8 WT, Trp53 ^−/− and Trp53 ^−/−;Pten ^−/− 1.15 cells grown in 2D as Round, Cobblestone, or Elongated. Heatmap, log₂ fold change, mean proportion across indicated lines. Grayscale heatmap, P‐values for each comparison. N = 2 independent experiments, four technical replicates/subline/experiment. Total cell number per condition, Table EV1.

11. L

    Representative images of cells quantified in (K). Cells classified by shape (Round, green; Cobblestone, red; Elongated, blue). Scale bar, 65 μm.

12. M

    Proliferation assay based on well confluence over time. N = 3 experiments set up with repeated cultures of each subline, 4–5 technical replicates/subline/experiment. Data are presented as mean ± SD. Unpaired, two‐tailed t‐test between WT and each of the sublines per time point. Significant P‐values annotated.

13. N

    Cell death assay, green object (Sytox green fluorescence) confluence over time. N = 2 experiments set up with repeated cultures of each subline, 4 technical replicates/subline in each experiment. Kruskal–Wallis ANOVA was performed at t = 10 h and t = 20 h, all comparisons are nonsignificant (P‐value > 0.05).

Figure 2. Loss of Pten is associated with collective invasion into ECM in a spheroid assay

1. A

   Schema, imaging of ID8 spheroids in three‐dimensional (3D) culture over time. Single cell suspensions were seeded onto and overlaid with ECM diluted in medium and then live‐imaged.

2. B

   Time series, showing a representative spheroid for each genotype, 12 h intervals. Arrowheads, protrusions into ECM. Scale bar, 20 μm. Right, cartoon of phenotype.

3. C

   Schema, analysis pathway to classify ID8 3D phenotypes. (1) Phase contrast images were segmented using CellProfiler. Shape, size, movement, texture, granularity and brightness measurements were extracted for each object. (2) Based on the measurements obtained from live imaging for each individual spheroid, we used CellProfiler Analyst and user‐supervised machine learning (FastGentle Boosting algorithm) to construct rules based on which the objects would be classified as “In Focus” or “Out‐of‐focus”. (3) The later were filtered out of the data set. (4) Additional machine learning was used to classify remaining ‘In‐focus’ objects as “Hyper‐protrusive” or “Spherical,” with high accuracy. (5) Data analysis pipeline was used to quantify the log₂ fold‐change of each phenotype relative to control for each subline.

4. D

   Frequency of Spherical and Hyper‐protrusive phenotypes in ID8 sublines, 6 h time intervals over 72 h. Heatmap (grayscale)—phenotype proportion (z‐score) in control (Wild‐type [WT]). Heatmap (blue‐red)—log₂ fold change from control. P‐values, bubble size (Cochran–Mantel–Haenszel test with Bonferroni adjustment). Black dot, homogenous effect across independent experiments (Breslow‐Day test, Bonferroni adjustment, non‐significant). N = 3 independent experiments, 3–5 technical replicates/experiment. Total spheroid number per condition, Table EV1.

5. E

   Representative phase contrast images of spheroids described in (D). Outlines pseudocoloured for classification (Spherical, green; Hyper‐protrusive, blue). Scale bars, 400 or 17 μm, as indicated. Magnified individual spheroids from boxed regions at indicated timepoints. Arrowheads, protrusions into ECM.

6. F

   Frequency of Spherical and Hyper‐protrusive phenotypes in ID8 parental spheroids expressing sgNT, sg2 Pten or sg5 Pten, 6‐h time intervals over 72 h. Heatmap (grayscale)—phenotype proportion (z‐score) in control (sgNT). Heatmap (blue‐red)—log₂ fold change from control. P‐values, bubble size (Cochran–Mantel–Haenszel test with Bonferroni adjustment). Black dot, homogenous effect across independent experiments (Breslow–Day test, Bonferroni adjustment, non‐significant). N = 3 independent experiments, 2–6 technical replicates/experiment. Total spheroid number per condition, Table EV1.

7. G

   Representative phase contrast images of spheroids described in (F). Outlines pseudocoloured for classification (Spherical, green; Hyper‐protrusive, blue). Magnified individual spheroids from boxed regions at indicated timepoints. Arrowheads, protrusions into ECM. Scale bars, 400 or 17 μm, as indicated.

8. H

   Schema, phenotypes of ID8 spheroids with analysed genotypes.

Source data are available online for this figure.

Figure EV2. Collective invasion into ECM in orthogonal assays

1. A

   Frequency of Spherical and Hyper‐protrusive phenotypes in ID8 WT, Trp53 ^−/−; Trp53 ^−/−;Pten ^−/− clones 1.12 and 1.15. spheroids, 6‐h time intervals over 72 h. Heatmap (grayscale)—phenotype proportion (z‐score) in control (WT). Heatmap (blue‐red)—log₂ fold change from control. P‐values, bubble size (Cochran–Mantel–Haenszel test with Bonferroni adjustment). Black dot, homogenous effect across independent experiments (Breslow–Day test, Bonferroni adjustment, nonsignificant). N = 3 independent experiments, 3–4 technical replicates/experiment. Total spheroid number per condition, Table EV1.

2. B

   Schema, 3D invasion into ECM of wounded ID8 monolayer.

3. C

   Representative images of 3D invasion assay, monolayers plated onto ECM, wounded and then overlaid with 50% ECM (gel). Outlines of invasive front at different time points, pseudocoloured by time (rainbow legend), overlaid as concatenate on phase image of initial wound. Boxes, regions for different timepoints. Yellow lines, initial wound. Scale bar, 400 or 45 μm (indicated).

4. D

   Quantification of (C). Graph, Relative Wound Density (RWD) at t _1/2 max (time when WT 50% closed). Data, mean (black square) ± SD for 3 independent experiments (large circles) with 3–6 technical replicates/subline/experiment (small circles). ANOVA with Tukey's HSD test; exact P‐values, annotated; ns, nonsignificant.

5. E

   Representative spider plots, leader cell movement in first 19 h of invasion of Trp53 ^−/− and Trp53 ^−/−;Pten ^−/− 1.15 ID8 cells. N = 2 independent experiments, 10–25 leader cells tracked in each, across 4–6 technical replicates/experiment.

6. F

   Confocal images of wounded monolayer invasive fronts, stained for F‐actin (Phalloidin). Arrowheads, protrusion tips. Scale bar, 65 μm. Representative of 7–12 fields imaged across n = 3 independent experiments, 4 technical replicates/experiment.

7. G

   Schema, loss of PTEN phenotype, a sheet‐like mode of invasion with most ECM‐abutting cells acting as “leader cells,” compared with leader and follower cell chains in WT.

Figure 3. PI3K‐AKT dependence of collective invasion

1. A

   Confocal images (single slice) of Trp53 ^−/− or Trp53 ^−/−;Pten ^−/− (1.15) spheroids expressing mNeonGreen‐tagged (mNG) biosensors for PI(4,5)P₂ (PH‐PLCδ1) or PIP₃ (CYTH3^2G/GRP1). Magnified images from boxed regions, max projection of 8 (PH‐PLCδ1) or 3 (CYTH3^2G/GRP1) z‐slices, pseudocoloured in FIRE LUT. Arrowheads: red, cell–cell contact; yellow, nucleus; green, protrusion tip. Scale bar, 7 μm. Representative of 8 (Trp53 ^−/−) or 10 (Trp53 ^−/−;Pten ^−/−) spheroids imaged across n = 2 independent experiments (PH‐PLCδ1) and 22 (Trp53 ^−/−) or 23 (Trp53 ^−/−;Pten ^−/−) spheroids imaged across n = 4 independent experiments (CYTH3^2G/GRP1).

2. B

   Intensity profiles for PH‐PLCδ1 and PH‐CYTH3 from spheroids shown in (A). Protrusions measured are annotated on images in upper panels, yellow lines. Arrowheads: red, protrusion tips.

3. C

   Schema, select PI‐kinases and phosphatases and their inhibitors participating in PIP₃ production and downstream AKT phosphorylation.

4. D

   Western blotting and quantitation for S6RP pS235/236, S6RP, GAPDH (sample integrity control) in Trp53 ^−/−;Pten ^−/− 1.15 spheroids treated with DMSO or inhibitors annotated in (B) for 2 days. Representative of n = 3 independent lysate preparations. Data, mean ± SD of pS235/236:total S6RP ratio, normalised to DMSO. P‐values, unpaired, two‐tailed t‐tests, as annotated.

5. E, F

   Quantitation of Trp53 ^−/−;Pten ^−/− 1.15 spheroids treated with DMSO, AKTi (AKT inhibitor II) or pan‐PI3Ki (LY294002), 6‐h time intervals over 72 h. (E) Heatmap (viridis)—area presented as mean of Z‐score values, normalised to control (DMSO). (F) Frequency of Spherical and Hyper‐protrusive phenotypes. Heatmap (grayscale)—phenotype proportion (z‐score) in control. Heatmap (blue‐red)—log₂ fold change from control. P‐values, bubble size (Cochran–Mantel–Haenszel test with Bonferroni adjustment). Black dot, homogenous effect across independent experiments (Breslow–Day test, Bonferroni adjustment, non‐significant). N = 2 independent experiments, 4–5 technical replicates/experiment. Total spheroid number per condition, Table EV1.

6. G

   Representative phase contrast images of spheroids described in (E). Outlines pseudocoloured for classification (Spherical, green; Hyper‐protrusive, blue). Magnified individual spheroids from boxed regions at indicated timepoints. Arrowheads, protrusions into ECM. Scale bar, 400 or 17 μm (indicated).

7. H, I

   Quantitation of ID8 Trp53 ^−/−;Pten ^−/− spheroids treated with PI3K isoform specific inhibitors: A66 (PI3Kα), AZD8186 (PI3Kβ), AS605240 (PI3Kγ) or CAL‐101 (PI3Kδ), 6‐h time intervals over 72 h. (H) Heatmap (viridis)—area presented as mean of Z‐score values, normalised to control (DMSO). (I) Frequency of Spherical and Hyper‐protrusive phenotypes. Heatmap (grayscale)—phenotype proportion (z‐score) in control. Heatmap (blue‐red)—log₂ fold change from control. P‐values, bubble size (Cochran–Mantel–Haenszel test with Bonferroni adjustment). Black dot, homogenous effect across independent experiments (Breslow–Day test, Bonferroni adjustment, non‐significant). N = 2 independent experiments, 3–5 technical replicates/experiment. Total spheroid number per condition, Table EV1.

8. J

   Representative phase contrast images of spheroids described in (H, I). Outlines pseudocoloured for classification (Spherical, green; Hyper‐protrusive, blue). Magnified individual spheroids from boxed regions at indicated timepoints. Arrowheads, protrusions into ECM. Scale bars, 400 or 17 μm, as indicated.

9. K

   Confocal image of Trp53 ^−/−;Pten ^−/− (1.15) spheroids stained for PI3Kβ (green), F‐actin (magenta) and Hoechst (grey). Magnified images from boxed regions, pseudocoloured in inverted grayscale (F‐actin) or FIRE LUT (PI3Kβ). Yellow or white/black arrowheads, enrichment of F‐actin or PI3Kβ at protrusion tips respectively. Scale bar, 5 μm. Representative of n = 5 spheroids.

10. L

    Intensity profiles for PI3Kβ (green) and F‐Actin (magenta) from spheroid in (K). Tip measured is annotated, ECM to body, yellow arrow, tip, arrowhead.

Source data are available online for this figure.

Figure EV3. Characterisation of Phosphoinositide enrichment in Invasion assays

1. A, C

   Confocal images, Trp53 ^−/− or Trp53 ^−/−;Pten ^−/− invasive monolayer fronts with cells expressing mNeonGreen (mNG) tagged biosensors for (A) PI(4,5)P₂ (PH‐PLCδ1) or (C) PIP₃ (CYTH3^2G). Representative of (A) 7 (Trp53 ^−/−) or 9 (Trp53 ^−/−;Pten ^−/−) fields or (C) 8 (Trp53 ^−/−) or 9 (Trp53 ^−/−;Pten ^−/−) fields imaged across n = 2 experiments set up with repeated cultures of each subline. Magnified boxed regions, pseudocoloured with FIRE LUT. Arrowheads: cell–cell contacts, black; protrusions, green; cell‐ECM contacts, red. Scale bar, (A)13 μm, (C) 6 μm.

2. B, D

   Intensity profiles for mNG PH‐PLCδ1 (B) or mNG PH‐ CYTH3^2G (D) from invasive monolayers on (A, C). Tips measured correspond to boxed, magnified regions on images in (A, C). Arrowhead, phosphoinositide‐rich region.

3. E

   Immunofluorescence and confocal imaging of Trp53 ^−/−;Pten ^−/− 1.15 spheroid stained for pS473‐AKT (green or FIRE LUT), F‐actin (magenta or black) and Hoechst (grey). Magnified images from boxed regions. Arrowheads, labelling of pS473‐AKT at protrusion tips. Scale bar, 5 μm. Representative of n = 5 spheroids.

4. F

   Intensity profile for pS473‐AKT (green) and F‐Actin (magenta) from spheroid in (A). Tip measured is annotated, ECM to body, yellow arrow; tip, white arrowhead.

Figure 4. The small GTPase ARF6 is required for Pten‐loss mediated ECM invasion

1. A, B

   Quantitation of ID8 Trp53 ^−/−;Pten ^−/− 1.15 spheroids expressing shScramble, shArf5 or shArf6, 6 h time intervals over 72 h. (A) Heatmap (viridis)—area presented as mean of Z‐score values, normalised to control (shScramble). (B) Frequency of Spherical and Hyper‐protrusive phenotypes. Heatmap (grayscale)—phenotype proportion (z‐score) in control. Heatmap (blue‐red)—log₂ fold change from control. P‐values, bubble size (Cochran–Mantel–Haenszel test with Bonferroni adjustment). Black dot, homogenous effect across independent experiments (Breslow–Day test, Bonferroni adjustment, non‐significant). N = 3 independent experiments, 4–5 technical replicates/experiment. Total spheroid number per condition, Table EV1.

2. C

   Representative phase contrast images of spheroids described in (A, B). Outlines pseudocoloured for classification (Spherical, green; Hyper‐protrusive, blue). Magnified individual spheroids from boxed regions at indicated timepoints. Arrowheads, protrusions into ECM. Scale bars, 400 or 17 μm, as indicated.

3. D–G

   Representative confocal images and intensity profiles of ID8 Trp53 ^−/− and Trp53 ^−/−;Pten ^−/− 1.15 cells expressing mNeonGreen (mNG)‐tagged ARF6 (green) and stained with Hoechst (grey) and F‐actin (magenta) at the invasive front of wounded monolayers (D, E) or in spheroids (max. projection ~ 10 Z‐slices) (F, G). Pseudo colour is FIRE LUT. Magnified images of boxed regions are shown. Arrowheads: cell–cell contacts, white; cell‐ECM contacts, blue; endosomes, grey; protrusion tips, yellow; intracellular pool, green. Scale bar, 5 μm. (E, G) Tips for which ARF6‐mNG and F‐Actin intensity profiles were measured are annotated, ECM to body, yellow arrow. (D) n = 2 independent experiments, 3–7 fields imaged/subline/experiment. (F) n = 3 independent experiments, 4–8 fields imaged/subline/experiment.

4. H

   Schema, mass spectrometry (MS) proteomic‐based TurboID approach for detecting ARF6‐proximal proteins.

5. I

   STRING network analysis of ARF6 interactions visualised using Cytoscape. Nodes manually annotated for known protein complexes. N = 4 independent lysate preparations from each subline.

6. J–L

   Heatmap, (J) unchanging, (K) Strong changing or (L) Weaker changing ARF6 interactors across genotypes. White to blue colour or blue to red, ARF6 interaction score, Log₂Fold Student's t‐test Difference in LFQ intensity compared to control ID8 Trp53 ^−/−;Pten ^−/− 1.15 TurboID alone. Interactors, sorted, descending order of mean interaction. Circle size, t‐test P‐value, coloured spots underneath denote the protein complex that each interactor belongs (in I), manual annotation. N = 4 independent lysate preparations from each subline.

Source data are available online for this figure.

Figure EV4. Further characterisation of ARF6 role upon Trp53 and Pten loss

1. A, B

   Western blot (A) and quantitation (B) of pS473‐AKT, AKT, ARF5, ARF6, GAPDH in ID8 Trp53 ^−/−;Pten ^−/− 1.15 cell lines expressing shScramble, shArf5 or shArf6. Representative blots of n = 3 independent lysate preparations. (B) Data, mean ± SD for ARF5, ARF6 and pS473‐AKT band intensity ratio, normalised to shScramble. P‐values, unpaired, two‐tailed t‐test; ns, not significant. GAPDH is loading control for all panels.

2. C

   Representative images, ID8 Trp53 ^−/−;Pten ^−/− 1.15 cell lines expressing shScramble, shArf5 or shArf6 in wounded monolayers invading ECM. Yellow lines, initial wound. Arrowheads, invasive protrusions. Outlines of invasive front pseudocoloured by time and overlaid as concatenate over phase image of initial wound. Scale bar, 45 μm. N = 3 independent experiments, 3–6 technical replicates/experiment.

3. D

   Quantitation of (C). Graph, Relative Wound Density (RWD) at t _1/2 max (time when shScramble 50% closed). Data, mean (black square) ± SD for 3 independent experiments (large circles), 3–6 technical replicates//experiment (small circles). P‐values, ANOVA with Tukey's HSD test; annotated when significant.

4. E, F

   Western blot (E) and quantitation (F) from pS473‐AKT, AKT, ARF6, VCL in ID8 Trp53 ^−/−;Pten ^−/− 1.15 cell lines expressing shScramble or shArf6 (5 individual shRNA sequences). Representative blots of n = 3 (ARF6) or n = 2 (pS473‐AKT and AKT) independent lysate preparations VCL is loading control for all panels. (F) Data, mean ± SD for ARF6 and pS473‐AKT band intensity normalised to shScramble. P‐values, unpaired, two‐tailed t‐tests; ns, not significant.

5. G

   Regression analysis. Scatter plot, mean Hyper‐protrusive level across all time points versus ARF6 protein levels (determined by western blot). Solid black line, best linear fit and dotted cyan lines, 95% confidence interval. P‐value and R ², annotated.

6. H

   Heatmap, Log₂‐transformed RNA‐sequencing read counts of each ARF GTPase in ID8 spheroids and 2D monolayers (Wild‐Type, WT [2D]) across n = 4 independent RNA preparations.

7. I

   Western blot and quantitation for ARF6 protein in ID8 sublines. VCL, loading control. Representative blots of n = 3 independent protein isolations. Quantitation, mean ± SD ARF6 intensity normalised to ID8 WT. P‐values, unpaired, two‐tailed t‐tests; ns, not significant.

8. J

   ARF6‐GTP levels in ID8 sublines. Normalised Optical Density (OD) of Arf6‐GTP G‐LISA. N = 3 independent lysate preparations, 3 technical replicates/experiment. Data, mean ± SD of independent replicates (large circles) with technical replicates shown (small circles). P‐values annotated, student's t‐test; ns, nonsignificant.

9. K

   Western blot, ARF6, T2A, V5, BFP and VCL from lysates extracted from TurboID (control) or ARF6‐TurboID‐expressing cell lines. VCL, loading control for T2A and BFP and sample integrity control for all other blots. N = 3 independent lysate preparations.

10. L

    Confocal images, ID8 Trp53 ^−/−;Pten ^−/− 1.15 cells expressing ARF6‐TurboID or TurboID, stained with T2A. Red box, cell–cell contacts, shown in higher magnification and pseudocoloured with FIRE LUT. Black arrowheads, cell–cell contact; green arrowheads, cell periphery. Scale, 10 μm. Representative images from two fields (ARF6‐TurboID) or three fields (TurboID alone) from one experiment.

11. M

    Western blot with Streptavidin HRP in ID8 Trp53 ^−/−;Pten ^−/− 1.15 cells treated with Biotin for at times and concentrations indicated. VCL was used as loading control. n = 1 lysate preparation.

12. N

    Gene Ontology Cell Compartment (GOCC) enrichment analysis of interactors identified in ID8 Trp53 ^−/−;Pten ^−/− 1.15 cells expressing ARF6‐TurboID compared to TurboID alone. Data, P‐value (−Log₁₀) of enrichment. N = 4 independent experiments. Red dotted line, significance threshold.

Figure EV5. Characterisation of Cytohesin and CYTH2 contribution to invasion

1. A

   Heatmap, log₂‐transformed RNA‐sequencing read counts of ARF GEFs in ID8 spheroids and 2D monolayers (Wild Type, WT [2D]) across n = 4 independent RNA preparations.

2. B, C

   Quantitation of ID8 Trp53 ^−/−;Pten ^−/− spheroids treated with 20 μΜ SecinH3, 6‐h time intervals over 72 h. (B) Heatmap (viridis)—area presented as mean of Z‐score values, normalised to control (DMSO). (C) Frequency of Spherical and Hyper‐protrusive phenotypes. Heatmap (grayscale)—phenotype proportion (z‐score) in control. Heatmap (blue‐red)—log₂ fold change from control. P‐values, bubble size (Cochran–Mantel–Haenszel test with Bonferroni adjustment). Black dot, homogenous effect across independent experiments (Breslow–Day test, Bonferroni adjustment, non‐significant). N = 3 independent experiments, 4–5 technical replicates/experiment. Total spheroid number per condition, Table EV1.

3. D

   Representative phase contrast images of spheroids described in (B, C). Outlines pseudocoloured for classification (Spherical, green; Hyper‐protrusive, blue). Magnified individual spheroids from boxed regions at indicated timepoints. Arrowheads, protrusions into ECM. Scale bars, 400 or 17 μm, as indicated.

4. E

   Representative images of ID8 Trp53^−/−;Pten^−/− 1.15 wounded monolayers treated with DMSO or SecinH3 (20 μΜ) in wounded monolayers invading ECM. Yellow lines, initial wound. Arrowheads, invasive protrusions. Outlines of invasive front pseudocoloured by time and overlaid as concatenate over phase image of initial wound. N = 3 independent experiments, 3–5 technical replicates/experiment. Scale bar, 200 or 45 μm.

5. F

   Quantitation of (E). Graph, Relative Wound Density (RWD) at t _1/2 max (time when DMSO 50% closed). Data, mean (black square) ± SD for 3 independent experiments (large circles), 3–6 technical replicates/experiment (small circles). Exact P‐value annotated, ANOVA with Tukey's HSD test.

6. G

   Spider plots of leader cell movement in the first 19 h of invasion of Trp53 ^−/−;Pten ^−/− 1.15 ID8 cells treated with DMSO or SecinH3. 10–25 leader cells were tracked per experiment (n = 2 set‐up using repeated cultures of each subline), across multiple technical replicates/experiment. Representative plots from cells tracked in one independent experiment shown.

7. H–P

   CYTH2 mRNA levels in (H, I) LCM normal ovarian surface epithelium versus HGSOC epithelium or normal ovarian stroma versus ovarian cancer‐associated stroma, or (J–P) bulk sequencing of normal ovary versus tumour. Specific data set, sample size (n) and P‐values (Mann–Whitney) annotated, whiskers Min‐Max, line at median.

8. Q–T

   Overall survival (% patients, months; TCGA OV data set), of patients grouped by low (M1) versus high (M2) levels based on a median split of (Q) CYTH2 mRNA, (R) CYTH2 exon 9.1 percentage spliced in ratio (PSI), (S) combination of ARF6 and CYTH2 mRNA, or (T) combination of ARF6 mRNA and CYTH2 Ex9.1 PSI. Median survival, sample size (n) and P‐value, Log‐rank test (Mantel‐Cox) annotated.

Figure 5. A functional proteomic CRISPR screen for ARF6‐proximal proteins controlling collective invasion

1. A

   Schema, (1) CRISPR screen. 26 ARF6‐proximal proteins from TurboID studies were investigated for their contribution to ARF6‐mediated invasion of ID8 Trp53^−/−;Pten ^−/− spheroids. (2) For each interactor, 5 sgRNAs were cloned into lentiviral CRISPR vectors. (3) A pooled approach was used, generating a KO cell line with all 5 sgRNAs (4) Live imaging performed. (5) Phenotype of each KO compared with nontargeting sgRNA.

2. B

   Frequency of Spherical and Hyper‐protrusive phenotypes upon pooled gRNA CRISPR of indicated targets (sorted based on hierarchical clustering) in ID8 Trp53 ^−/−;Pten ^−/− clone 1.15 cells, performed in four parts (Iterations indicated). Heatmap (grayscale)—phenotype proportion (z‐score) in control (sgNT). Heatmap (blue‐red)—log₂ fold change from control. P‐values, bubble size (Cochran–Mantel–Haenszel test with Bonferroni adjustment). Black dot, homogenous effect across independent experiments (Breslow–Day test, Bonferroni adjustment, nonsignificant). N = 3–4 independent experiments, 3–6 technical replicates/experiment. Total spheroid number per condition, Table EV1.

3. C

   Western blot, β1‐integrin (ITGB1), pS473‐AKT, AKT, ARF6 from deconvolved ITGB1 sgRNA‐expressing cells. VCL, loading control for ITGB1, sample integrity control for other blots. Representative blots of n = 3 independent lysate preparations.

4. D

   Quantitation of (C). Data, mean ± SD for pS473‐AKT:total AKT band intensity ratio, total AKT or ARF6 intensity, normalised to control (sgNT ID8 Trp53 ^−/−;Pten ^−/− clone 1.15) cells. P‐values, unpaired, two‐tailed t‐test.

5. E, F

   Frequency of Spherical and Hyper‐protrusive phenotypes in ID8 Trp53 ^−/−;Pten ^−/− 1.15spheroids upon CRISPR‐mediated KO of (E) Itgβ1 or (F) Agap1, 6 h time intervals over 72 h. Heatmap (grayscale)—phenotype proportion (z‐score) in control (sgNT). Heatmap (blue‐red)—log₂ fold change from control. P‐values, bubble size (Cochran–Mantel–Haenszel test with Bonferroni adjustment). Black dot, homogenous effect across independent experiments (Breslow–Day test, Bonferroni adjustment, non‐significant). N = 3 independent experiments, 1–5 technical replicates/experiment. Total spheroid number per condition, Table EV1.

6. G

   Representative phase contrast images of spheroids described in (E, F). Outlines pseudocoloured for classification (Spherical, green; Hyper‐protrusive, blue) at indicated timepoints. Magnified individual spheroids from boxed regions. Arrowheads, protrusions into ECM. Scale bars, 400 or 17 μm, as indicated.

7. H

   Representative confocal images of Trp53 ^−/− and Trp53 ^−/−;Pten ^−/− clone 1.15 spheroids expressing sgNT, sgAgap1 (sg3) or sgItgb1 (sg4), stained for collagen IV (grayscale) and F‐Actin (magenta). Boxed areas, basement membrane region in higher magnification. Arrowheads, Collagen IV labelling that is: well‐defined, green; fragmented, yellow; absent, navy. Scale bar, 53 μm.

8. I

   Quantitation of (H). Collagen IV basement membrane staining as Defined, Fragmented, or Absent in spheroids set up across n = 3 independent experiments, 1 technical replicate/experiment, 5–9 fields imaged per technical replicate, 365 spheroids scored in total. Data, mean ± SD of % of spheroids in each phenotype for independent experiments, with circles representing technical replicates. Unpaired t‐test, P‐values annotated.

Source data are available online for this figure.

Figure 6. The ARFGAP AGAP1 controls invasion and stratifies survival

1. A

   Schema, AGAP1 isoform domains. GLD, GTP binding‐like domain; PH, Pleckstrin homology; ANK, Ankyrin; ARF GAP, ARF GTPase‐activating Protein. Based on information found in www.ensembl.org (Cunningham et al, ; ‘Long’ isoform, Transcript ID: ENST00000304032.13 for the human genome, and ENSMUST00000027521.15 for the mouse genome) or 804 amino acids (‘Short’ isoform, Transcript ID: ENST00000336665.9 for the human and ENSMUST00000190096.7 for the mouse genome) and previously described annotations of AGAP1 domains (Nie et al, 2002).

2. B

   Heatmap, differential association of isoforms with phospholipids. Data, Log₂‐transformed % of total signal between AGAP1‐S versus AGAP1‐L GST‐tagged PH domain association with each phospholipid. P‐value, circles size (unpaired t‐test). n = 3 blots per condition.

3. C

   Western blots of ID8 Trp53^−/−;Pten ^−/− 1.15 cells expressing either sgNT or sgAgap1 (sg3) and either mNeonGreen (mNG) or CRISPR‐resistant mNG‐Agap1_S or ‐L isoforms. Blotted for ARF6, pS473‐AKT, AKT, mNG, and VCL. VCL is loading control for AKT, pS473‐AKT and ARF6 and sample integrity control for others. n = 3 independent lysate preparations.

4. D

   Quantitation of (C). Data, mean ± SD for ARF6 and pS473/AKT band intensity ratio, normalised to sgNT. P‐values, unpaired two‐tailed t‐test, annotated when significant.

5. E, F

   Quantitation of ID8 Trp53 ^−/−;Pten ^−/− 1.15 spheroids treated with sgNT or AGAP1‐targeting sg3 and expressing either mNG or mNG‐fusion with either isoform of AGAP1, 6 h time intervals over 72 h. (E) Heatmap (viridis)—area presented as mean of Z‐score values, normalised to control (sgNT). (F) Frequency of Spherical and Hyper‐protrusive phenotypes. Heatmap (grayscale)—phenotype proportion (z‐score) in control. Heatmap (blue‐red)—log₂ fold change from control. P‐values, bubble size (Cochran–Mantel–Haenszel test with Bonferroni adjustment). Black dot, homogenous effect across independent experiments (Breslow‐Day test, Bonferroni adjustment, nonsignificant). N = 3 independent experiments, 5–6 technical replicates/experiment. Total spheroid number per condition, Table EV1.

6. G

   Representative phase contrast images of spheroids described in (E, F). Outlines pseudocoloured for classification (Spherical, green; Hyper‐protrusive, blue). Magnified individual spheroids from boxed regions at indicated timepoints. Arrowheads, protrusions into ECM. Scale bars, 400 or 17 μm, as indicated.

7. H–J

   ARF6 and AGAP1 mRNA levels in LCM normal ovarian surface epithelium versus HGSOC epithelium or normal ovarian stroma versus OC‐associated stroma. Specific data sets, sample size (n) and P‐values (Mann–Whitney) annotated, whiskers Min–Max, line at median.

8. K–O

   Overall survival (% patients, months; TCGA OV data set), of patients grouped by low (M1) versus high (M2) levels, based on a median split, of (K) ARF6 mRNA, (L) AGAP1 mRNA, (M) AGAP1 Exon 14 percentage spliced in ratio (PSI), (N) combination of ARF6 and AGAP1 mRNA, or (O) combination of ARF6 mRNA and AGAP1 Ex14 PSI. Median survival, sample size (n) and P‐value, Log‐rank test (Mantel‐Cox) annotated.

Source data are available online for this figure.

Figure 7. ARF6 controls invasion by regulating recycling of active integrins

1. A, B

   Immunofluorescence and confocal imaging of Trp53 ^−/−;Pten ^−/− 1.15 spheroids stained for α5‐integrin or β1‐integrin (grey or FIRE LUT), Hoechst (blue) and F‐actin (magenta). Magnified images from boxed regions shown. Arrowheads, labelling at protrusion tips. Scale bars, 5 μm. Representative of n = 3 spheroids imaged. (B) Intensity profiles for integrins (grey) and F‐actin (magenta) from spheroids in (A). Tip measured is annotated, ECM to body, yellow arrow, tip, white arrowhead.

2. C, D

   Immunofluorescence and confocal imaging of Trp53 ^−/−;Pten ^−/− 1.15 spheroids stained for pFAK (Y379) or pSRC Family Kinases (SFK pY416) (grey or FIRE LUT), Hoechst (blue) and F‐actin (magenta). Magnified images from boxed regions shown. Arrowheads, positive staining. Scale bars, 5 μm. Representative of n = 5 spheroids imaged. (D) Intensity profiles for active FAK and Src (grey) and F‐actin (magenta) from spheroids in (C). Tip measured is annotated, ECM to body, yellow arrow, tip, white arrowhead.

3. E–H

   Representative capture ELISA graphs (E, G) and associated quantitation (F, H) for recycling of internalised cargoes between Trp53 ^−/− versus Trp53 ^−/−;Pten ^−/− cells or Trp53 ^−/−;Pten ^−/− cells expressing shScramble versus shArf6 for active β1‐integrin. Graphs shown are representative of n = 2 (E) or n = 3 (G) independent replicates. Data, mean (black square) ± SD for repeated experiments (large circles), 1–3 technical replicates/experiment/timepoint (small circles), two‐tailed t‐test, P‐values are annotated.

4. I–K

   Overall survival (% patients, months; TCGA OV data set) of patients grouped into combined expression based on median mRNA split. (I) Low (red line, M1) or high (blue line, M2) expression for all mRNA, control, remaining patients (green line), (J), same as (I), but CYTH2 Ex9 PSI, rather than total CYTH2. (K), as for (I), but PTEN protein levels split by quantiles (red and blue, Q1 + Q2, Q3, low PTEN, green Q4, high PTEN). Median survival, sample size (n) and P‐value, Log‐rank test (Mantel‐Cox) annotated.

5. L

   Differential abundance (x, Log Ratio between conditions; y, Log₁₀ q‐values) of proteins in PIP₃‐responsive module (ARF6^HI‐AGAP1^HI‐CYTH2^2G) versus PI(4,5)P₂‐responsive ARF module (ARF6^HI‐AGAP1^HI‐CYTH2^3G) protein samples. Reverse Phase Protein Array Data, TCGA OV. Significantly altered components in AKT signalling pathway labelled (−Log₁₀ q‐value > 1.3).

6. M

   Schema, molecular model for ARF GTPase regulation of integrin‐dependent invasion.

Source data are available online for this figure.