A RAB5/RAB4 recycling circuitry induces a proteolytic invasive program and promotes tumor dissemination - PubMed (original) (raw)
. 2014 Jul 21;206(2):307-28.
doi: 10.1083/jcb.201403127.
Andrea Palamidessi 1, Paola Marighetti 1, Stefano Confalonieri 2, Fabrizio Bianchi 3, Chiara Malinverno 1, Giovanni Mazzarol 3, Giuseppe Viale 3, Ines Martin-Padura 3, Massimilliano Garré 1, Dario Parazzoli 1, Valentina Mattei 3, Salvatore Cortellino 1, Giovanni Bertalot 3, Pier Paolo Di Fiore 4, Giorgio Scita 5
Affiliations
- PMID: 25049275
- PMCID: PMC4107781
- DOI: 10.1083/jcb.201403127
A RAB5/RAB4 recycling circuitry induces a proteolytic invasive program and promotes tumor dissemination
Emanuela Frittoli et al. J Cell Biol. 2014.
Abstract
The mechanisms by which tumor cells metastasize and the role of endocytic proteins in this process are not well understood. We report that overexpression of the GTPase RAB5A, a master regulator of endocytosis, is predictive of aggressive behavior and metastatic ability in human breast cancers. RAB5A is necessary and sufficient to promote local invasion and distant dissemination of various mammary and nonmammary tumor cell lines, and this prometastatic behavior is associated with increased intratumoral cell motility. Specifically, RAB5A is necessary for the formation of invadosomes, membrane protrusions specialized in extracellular matrix (ECM) degradation. RAB5A promotes RAB4- and RABENOSYN-5-dependent endo/exocytic cycles (EECs) of critical cargos (membrane-type 1 matrix metalloprotease [MT1-MMP] and β3 integrin) required for invadosome formation in response to motogenic stimuli. This trafficking circuitry is necessary for spatially localized hepatocyte growth factor (HGF)/MET signaling that drives invasive, proteolysis-dependent chemotaxis in vitro and for conversion of ductal carcinoma in situ to invasive ductal carcinoma in vivo. Thus, RAB5A/RAB4 EECs promote tumor dissemination by controlling a proteolytic, mesenchymal invasive program.
© 2014 Frittoli et al.
Figures
Figure 1.
RAB5A predicts poor clinical outcome in breast cancer patients and its expression is elevated in breast cancer lymph node metastases. (A–D) Kaplan–Meier plots for breast cancer patients stratified by RAB5A expression for the following groups of patients: (A) all 980 patients; (B) 695 patients with lymph node negative (N0) breast tumors; (C) 528 patients with lymph node–negative and ER-positive (N0 ER+) breast tumors; (D) 166 patients with lymph node–negative, ER-positive, and Grade 2 (N0 ER+ G2) breast tumors. High RAB5A, >75th percentile; low RAB5A, ≤75th percentile. Y axes = the probability of local and distant relapse free survival (DFS). X axes = years of follow up. P-values are from a log-rank test. (E) RAB5A prognostic significance is independent from Ki67, and its expression is not associated with a proliferation signature. (E, left) Multivariate analysis with a Cox proportional hazard model on prognosis of patients with breast cancer. The RAB5A and Ki67 genes were used as covariates. High RAB5A or Ki67 expression levels, >75th percentile of the normalize expression distribution. Low RAB5A or Ki67 expression levels, ≤75th percentile. HR, hazard ratio by Cox model. 95% CI, 95% confidence intervals of HR. P-value, likelihood ratio χ2 test on the null hypothesis that the parameter estimate for the RAB5A high/low, or Ki67 high/low, covariate is zero. (E, right) Gene set enrichment analysis (GSEA) analysis of a known proliferation signature (Ben-Porath et al., 2008) among genes ranked by signal-to-noise metric based on their correlation with Ki67 high/low or RAB5A high/low. FDR, false discovery rate based on 1,000 random permutations of class labels. (F) IHC staining of RAB5 on FFPE samples of a human primary infiltrating breast tumor (PT) and matched nodal metastasis (NM). N, normal duct; L, normal negative lymphocytes. Boxed regions are enlarged below. Bar, 500 µm. (G) Percentage of low and high RAB5A-expressing primary tumors and metastases. Normal breast tissue scores always range between 0 and 1. **, P < 0.000271 (χ2 test).
Figure 2.
RAB5A is required for tumor dissemination and for the conversion from DCIS to infiltrating mammary carcinoma. (A and B) Doxycycline-inducible RAB5AS34N- and control MDA-MB-231 cells were injected into the mammary fat pads of NSG mice. After 3 wk, mice were fed with doxycycline. Metastases were analyzed 4 wk after removal of primary tumor. (A) Ipsilateral metastasis (arrowhead) of control or RAB5AS34N-MDA-MB-231 tumors. (A, right) Quantification (mean ± SEM [error bars]; n = 10 mice/group) of the number and size of disseminated tumors. (B) H&E of control and RAB5AS34N-MDA-MB-231 lung tissue sections (left). The size of metastatic nodules in lungs is the mean tumor area/lung area ± SEM (error bars; n = 9 mice/group repeated in three independent experiments). (C) Mixtures of CFSE-labeled (green) scramble-transfected control (Ctr) and eFluor 670–labeled (red) RAB5A,B,C-depleted MDA-MB-231 cells (0.5 × 106 each) were coinjected into the tail vein of NSG mice. (C, left) Images of lung tissue. (C, middle) The number (mean ± SEM [error bars]; n = 10) of labeled control and RAB5A,B,C-depleted cells/field of lung tissue at the indicated times. **, P < 0.005. (C, right) Efficacy of RAB5 gene silencing by QRT-PCR. **, P < 0.005. (D) Doxycycline-inducible RAB5AS34N- and control-MCF10.DCIS.com cells were injected subcutaneously into NSG mice. After 4 d, mice were fed with doxycycline. (D, left) IHC analyses of control (Ctr) and RAB5AS34N-MCF10.DCIS.com xenografts performed at 1 and 3 wk after doxycycline treatment. Representative images from three independent experiments are shown (n = 5 mice/experimental condition). (D, right) mRNA level of RAB5AS34N expression in control (Ctr) and RAB5AS34N tumors. Error bars indicate SEM. **, P < 0.005. (E) Control (Ctr) and RAB5A-MCF10.DCIS.com cells (5 × 105 each) were labeled with eFluor 670 and injected into NSG mice tail vein. (E, left) Images of lung tissue. (E, middle) The number (mean ± SEM [error bars]; n = 10) of labeled control and RAB5A-MCF10.DCIS.com cells per field at the indicated times. **, P < 0.005. (E, right) RAB5A expression by immunoblotting. Bars: (B) 2 µm; (C) 80 µm; (D) 100 µm; (E) 80 µm.
Figure 3.
RAB5A is sufficient to promote intratumoral cell motility and distant dissemination. (A) GFP-LifeAct control (Ctr) or GFP-LifeAct-RAB5A-(RAB5A) HeLa cells were injected into the mammary fat pads of NSG mice. Metastases were analyzed 4 wk after removal of primary tumor. (A, left) Ipsilateral metastasis (arrowheads) of control or RAB5A-HeLa tumors. (A, right) Quantitation of number and volume (mean ± SEM [error bars]; n = 10 mice/group) of disseminated tumors nodules. **, P < 0.005. (B, left) H&E and anti-GFP staining of FFPE lung tissue sections. Magnified boxed regions show metastasis. (B, right) The size of metastatic nodules is the mean tumor area/total lung area ± SEM (error bars; n = 9 mice/group repeated in two independent experiments). **, P < 0.005. (C) Tumors from GFP-LifeAct control (Ctr) or GFP-LifeAct-RAB5A-(RAB5A)-HeLa cells injected into the mammary fat pads of NSG mice were analyzed by two-photon microscopy. Green, GFP-LifeAct; gray, collagen structure (SHG). (D) Tumor invasive front visualized: in the top panels by projecting ∼40 serial z sections (green, GFP-LifeAct; gray, collagen structure [SHG]; or, in bottom panels, by IHC. (E) Intratumoral motion analysis of control and RAB5A-HeLa cells was obtained by overlaying 10 differentially colored, consecutive frames of time-lapse recording (
Videos 1 and 2
; left). Coloring indicates motile cells. The percentage of motility events/field of view/tumor is the mean ± SEM (error bars; n = 45). Bars: (B, left) 2 mm; (B, right) 0.2 mm; (C) 400 µm; (D, top) 200 µm; (D, bottom) 10 µm; (E) 80 µm.
Figure 4.
RAB5A is necessary and sufficient to promote matrix-metalloprotease–dependent invasion into a 3D matrix. (A) RAB5A promotes organotypic cell invasion and remodels stromal collagen. (A, left) H&E-stained cross-sections of GFP-LifeAct control (Ctr) or GFP-LifeAct-RAB5A-(RAB5A) HeLa cells grown for 14 d on organotypic collagen matrix preconditioned with U937 macrophages, in the absence or presence of GM6001. Bar, 400 µm. (B) Quantification of the number of invading cells/condition, with respect to the number of invading RAB5A-HeLa cells, is the mean invasion index ± SEM (error bars) of four independent experiments (n = 45). *, P < 0.01. (C and D) Doxycycline-induced control- and RAB5AS34N-MDA-MB-231 (C) or RAB5AS34N-MCF10.DCIS.com (D) cells were assessed for their invasiveness by placing them on one side of a chamber slide in which 2.3 mg/ml acid extracted–only polymerized type I collagen gel and 100 ng/ml HGF were added. Examples of cell migration tracks are shown on the left (
Videos 5 and 6
). (C and D, right) Cell invasion is expressed as the mean forward invasion index ± SEM (error bars; n = 75 single cells/experiment repeated in four independent assays). *, P < 0.01; **, P < 0.005. (E) Doxycycline-inducible control (empty) or RAB5AS34N-MF10.DCIS.com cells were grown on a thick 1:1 Matrigel/type I collagen mixture and overlaid with diluted Matrigel (2 mg/ml). After 1 wk, cells were treated with doxycycline and/or HGF (20 ng/ml), or were mock treated. Bars, 400 µm. (E, right) The percentage of structures with invasive outgrowths was expressed as mean ± SEM (error bars) of four independent experiments; n = 35. **, P < 0.005.
Figure 5.
RAB5A is essential for HGF-induced invadosome formation. (A and B) Doxycycline-induced control and RAB5AS34N-MDA-MB-231 (A) or RAB5AS34N-MCF10.DCIS.com (B) cells plated onto fluorescently conjugated gelatin were stimulated with HGF (100 ng/ml) for 3 h. (B, left) Images of cells stained with phalloidin (left), fluorescently conjugated gelatin (middle), and merged channels (right). (B, right) Quantification of gelatin degradation was expressed as a degradation index (calculated as described in Materials and methods). Data are the mean ± SEM (error bars; n = 70 cells/experiment in five independent ones). **, P < 0.005. (C, left) TIRF microscopy of RAB5A-HeLa cells. F-actin and RAB5A were detected with phalloidin and anti–RAB5 Ab (RAB5), respectively. (C, right) xz sections of control and RAB5A-HeLa cells plated onto fluorescently conjugated gelatin (red) and stained with phalloidin (green). (D) Control and RAB5A-HeLa cells were plated onto fluorescently conjugated gelatin (middle) overnight under serum-starved conditions and stimulated with suboptimal doses of HGF (1 ng/ml). F-actin and RAB5A were detected with phalloidin and anti-RAB5A Ab (RAB5A), respectively. Insets show magnifications of the boxed regions. Arrows indicate invadosomes. (D, right) Gelatin degradation was expressed as a degradation index (see Materials and methods). Data are the mean ± SEM (n = 50 cells/experiment in four independent ones). **, P < 0.005. Bars: (A–C) 20 µm; (D) 10 µm.
Figure 6.
The RAB4–RABENOSYN-5 recycling pathway is necessary for HGF-induced invadosome formation. (A) Serum-starved MCF10.DCIS.com cells, transfected with scrambled siRNA (Ctr), siRNAs against RABENOSYN-5 (siRNA RBNS-5) or RAB4A and -B (siRNA RAB4), or RABAPTIN-5. GFP-RAB4AS22N-MCF10.DCIS.com cells were induced or not induced with doxycycline and serum starved. Cells were plated onto fluorescently conjugated gelatin (green), stimulated with HGF for 3 h, or left in serum free conditions (SF), and stained with phalloidin (red). Bar, 15 µm. (A, left) GFP-RAB4AS22N (GFP-RAB4A) expression was verified by immunoblotting. (B) Gelatin degradation was expressed as a degradation index. Data are the mean ± SEM (error bars; n = 75 cells/experiment in four independent ones). *, P < 0.01. Silencing of RAB4A, RAB4B, and RABENOSYN-5 genes was performed using two independent siRNA oligos, which gave identical results (not depicted), and was verified by QRT-PCR (right). **, P < 0.005. (C) MCF10.DCIS.com cells were cotransfected with scrambled siRNA (Ctr) or siRNAs against RABENOSYN-5 (siRNA RBNS-5) together with siRNA-resistant RABENOSYN-5 wild type (resWT-RBNS-5) or mutants impaired in binding to RAB4 (resRAB4-Δ-RBNS-5) or to RAB5 (resRAB5-Δ-RBNS-5), or to both GTPases (resRAB4/5-Δ-RBNS-5) fused to GFP. Cells were plated onto fluorescently conjugated gelatin (red), stimulated with HGF for 3 h, or left in serum free conditions, and stained with DAPI (blue). Arrows, matrix degradation areas. Bar, 20 µm. (D) Gelatin degradation is expressed as a degradation index, as indicated in B. Data are the mean ± SEM (error bars; n = 65 cells/experiment in three independent ones). *, P < 0.01. Silencing of RABENOSYN-5 was verified by QRT-PCR (right). **, P < 0.005.
Figure 7.
HGF-induced fast recycling of MT1-MMP depends onto RAB5, RAB4, and RABENOSYN-5. (A) Cherry-MT1-MMP–expressing HeLa cells were transfected with scrambled siRNA or siRNAs against RAB5A,B,C (siRNA-RAB5) or RAB4A and -B (siRNA RAB4 A/B), RABENOSYN-5 (siRNA RBNS-5), or integrin β3 (siRNA β3). Serum-starved cells were incubated with anti–MT1-MMP antibody at 16°C for 2 h. After a mild acid wash to remove surface antibody, cells were switched to 37°C and stimulated with HGF (100 ng/ml), or left in serum free (SF) conditions. At the indicated time points, cells were fixed, and stained in the absence of permeabilization with FITC-conjugated secondary antibody (green). Cherry-MT1-MMP (red) was detected by epifluorescence. The relative cell surface levels of MT1-MMP were quantified using ImageJ software on nonsaturated images, and expressed as relative cell surface MT1-MMP levels with respect to HGF-stimulated control cells after 15 min (set at 100%). Data are the mean ± SEM (error bars; n = 25 cells repeated in three independent experiments). (B) Steady-state cell surface levels of MT1-MMP. Cherry-MT1-MMP-HeLa cells were transfected with scrambled siRNA or siRNAs against RAB5A,B,C (siRNA-RAB5) or RAB4A and -B (siRNA RAB4 A/B), RABENOSYN-5 (siRNA RBNS-5), or integrin β3 (siRNA β3). Cells were incubated with anti–MT1-MMP antibody at 4°C for 2 h, washed, fixed, and stained in the absence of permeabilization with FITC-conjugated secondary antibody (green). (C) Silencing of RAB5 impairs MT1-MMP internalization. Cherry-MT1-MMP-HeLa cells were transfected with scrambled (Ctr) or anti-RAB5A,B,C siRNAs. Serum-starved cells were incubated with anti–MT1-MMP antibody at 16°C for 2 h. Cells were washed, fixed, permeabilized with 0.1% Triton X-100, and stained with FITC-conjugated secondary antibody (green). Bottom right, efficacy of gene silencing by QRT-PCR. Data are the mean ± SEM (error bars). **, P < 0.005. Bars, 10 µm.
Figure 8.
Genetic and functional interference with αVβ3, but not α5β1, integrin inhibits RAB5A- and HGF-induced invadosome formation and matrix degradation. (A) Confocal analysis of RAB5A-HeLa cells stained with phalloidin (red) or the indicated Ab (green). (B) Still images from TIRF microscopy time-lapse of HeLa cells transfected with RAB5, GFP-β3, and RFP-LifeAct (see also
Video 8
). (C) HeLa cells cotransfected with RAB5A and MT1-MMP were subjected to PLA (red dots) with abs against MT1-MMP and integrin β3 or stained with phalloidin (green). Negative controls (Ctr) were cells stained with oligonucleotide-labeled PLA probes alone. Confocal images in A, B, and C are representative of >100 analyzed cells. (D) HeLa cells were transfected with scrambled (Ctr), anti-β1, or anti-β3 siRNA, or incubated with the inhibitory antibodies, 4B4 and LM609, against β1 and αVβ3, respectively. Serum-starved cells were plated onto fluorescently conjugated gelatin (red), stimulated with HGF (100 ng/ml) for 3 h, or left in serum free conditions (SF), and stained with phalloidin (green). Insets are magnifications of boxed regions. Quantification of gelatin degradation was expressed as a degradation index (relative to the area of degradation of control, HGF-stimulated cells normalized for cell number). Data are the mean ± SEM (error bars; n = 60 cells/experiment in five independent experiments). Two independent siRNAs were used for each integrin with similar results. Silencing of β1 and β3 genes was verified by QRT-PCR (right). **, P < 0.005. Arrows in A–D indicate examples of invadosomes. (E) MDA-MB-231 and MCF10.DCIS.com cells were transfected with scrambled (Ctr) or anti-β3 siRNA. Serum-starved cells were then plated onto fluorescently conjugated gelatin (red), stimulated with HGF (100 ng/ml) for 3 h, or left in serum free conditions (SF), and stained with phalloidin (green). Gelatin degradation is expressed as a degradation index. Data are the mean ± SEM (error bars; n = 40 cells/experiment in three independent ones). Two independent siRNAs were used for each integrin with similar results. Silencing β3 genes was verified by QRT-PCR (right). **, P < 0.005. Bars: (A) 15 µm; (B–E) 20 µm.
Figure 9.
The RAB5A/RAB4A circuitry promotes 3D matrix invasion in an MT1-MMP– and β3 integrin–dependent manner, and is dysregulated in invasive breast cancer. (A) RAB5-HeLa cells were transfected with the indicated siRNAs or scrambled control (Ctr), and tested for invasion as described in Fig. 4 (C and D). (A, top) Examples of cell migration tracks (see
Video 9
). (A, bottom) Quantification of cell invasion is the mean forward invasion index ± SEM (error bars; n = 60 cells/experiment/siRNA repeated in four independent assays). *, P < 0.05; **, P < 0.005. Gene silencing was verified by QRT-PCR (bottom right). **, P < 0.001. (B) MDA-MB-231 cells were transfected with the indicated siRNAs or scrambled control (Ctr) and tested for invasion as described in Fig. 4 (C and D). (B, left) Examples of cell migration tracks (see
Video 10
). Quantification of cell invasion is expressed (right) as the mean forward invasion index ± SEM (error bars; n = 40 cells/experiment/siRNA repeated in three independent invasion assays). *, P < 0.05; **, P < 0.005. Silencing of the different genes was verified by QRT-PCR (right). **, P < 0.001. (C) Control and RAB4AS22N-MCF10.DCIS.com cells were induced with doxycycline and tested for invasion as described in Fig. 4 (C and D). (C, left) Examples of cell migration tracks (Video 10). (C, right) Quantification of cell invasion is the mean forward invasion index ± SEM (error bars; n = 45 cells/experiments in two independent experiments). (D) Doxycycline-inducible RAB4AS22N- and control-MCF10.DCIS.com cells were injected subcutaneously into NSG mice. After 2 d, mice were fed with doxycycline. Histological (H&E) and IHC analyses with anti–α smooth muscle actin (SMA) of control (Ctr) and RAB4AS22-MCF10.DCIS.com xenografts were performed at 1 and 3 wk after doxycycline treatment. Images are from three independent experiments (n = 6 mice/experimental condition). Bars, 100 µm.
Figure 10.
RAB4A is amplified in invasive breast cancer. (A) RAB4A is amplified in various tumors. The cBio Cancer Genomics Portal (
http://www.cbioportal.org/public-portal/
) was queried for RAB4A across various tumor datasets. The alteration frequency, type of alterations of RAB4, availability of mutation analysis, and Copy Number Alteration (CNA) in the various tumor cohorts are shown. (B) RAB4A is, among the RAB family members, the most frequently amplified gene in invasive breast carcinoma. The percentage of each RAB gene amplification in invasive breast carcinoma reported in the Cancer Genome Atlas (TCGA) is shown. Only the genes with a frequency of amplification >5% are shown. (C) RAB4A is overexpressed in breast cancer. Shown are examples of the data of
Table S2
. Circled areas are enlarged below. Bars, 100 µm.
References
Publication types
MeSH terms
Substances
LinkOut - more resources
Full Text Sources
Other Literature Sources
Medical
Miscellaneous