Endosomal sorting and c-Cbl targeting of paxillin to autophagosomes regulate cell-matrix adhesion turnover in human breast cancer cells

Post-translational mechanisms regulating cell-matrix adhesion turnover during cell locomotion are not fully elucidated. In this study, we uncovered an essential role of Y118 site-specific tyrosine phosphorylation of paxillin, an adapter protein of focal adhesion complexes, in paxillin recruitment to autophagosomes to trigger turnover of peripheral focal adhesions in human breast cancer cells. We demonstrate that the Rab-7 GTPase is a key upstream regulator of late endosomal sorting of tyrosine118-phosphorylated paxillin, which is subsequently recruited to autophagosomes via the cargo receptor c-Cbl. Essentially, this recruitment involves a direct and selective interaction between Y118-phospho-paxillin, c-Cbl, and LC3 and is independent from c-Cbl E3 ubiquitin ligase activity. Interference with the Rab7-paxillin-autophagy regulatory network using genetic and pharmacological approaches greatly impacted focal adhesion stability, cell locomotion and progression to metastasis using a panel of human breast cancer cells. Together, these results provide novel insights into the requirement of phospho-site specific post-translational mechanism of paxillin for autophagy targeting to regulate cell-matrix adhesion turnover and cell locomotion in breast cancer cells.


INTRODUCTION
Cell migration is a physiological process fundamental for embryonic development, immune and inflammatory responses, wound healing, and tissue homeostasis [1,2]. Aberrant cell migration has been associated with several pathologies including rare inherent diseases such periventricular heterotopia [3], Baraitser-Winter syndrome [4] and more common diseases such as cancer progression to metastasis [5].
In general, motile cells exhibit finger-like protrusions (e.g. lamellipodia, filopodia) of the plasma membrane that are stabilized by the formation of focal adhesions (FAs). The synchronous cycle of FA assembly/disassembly are essential for cell locomotion [6] and require the integration of multiple signals from extracellular matrixreceptor interactions, cell cytoskeleton and binding to intracellular proteins primarily those involved in endocytic trafficking, in particular members of the Rab GTPase family involved in early endosome development [7][8][9][10][11], many of which are amplified in invasive cancers and have been associated with enhanced cancer cell invasiveness [12][13][14]. However, there remains a fundamental gap in understanding the molecular mechanisms by which Rab GTPases that are involved in late endosomal maturation regulate FA turnover, with a particular focus on FA posttranslational modifications that trigger these events.
In this study, we report a novel role for the late endosomal protein Rab7 GTPase in the regulation of the selective recruitment of tyrosine 118 (Y118)phosphorylated paxillin, but not Y31-phospho-paxillin to autophagosomes via its interaction with c-Cbl, which serves as a cargo receptor independently of its E3 ubiquitin ligase activity, in human breast cancer cells. This novel Rab7-mediated turnover of 118Y-p-paxillin was inhibited upon Atg12 downregulation, thereby reinforcing the implication of the autophagic pathway. This regulatory axis was not seen with other FA proteins

Research Paper
Oncotarget 31200 www.impactjournals.com/oncotarget such as FAK and was not affected by Src manipulation. Although autophagy-mediated FA turnover has recently been identified in mouse mammary 4T1 cells [15], studies describing this mechanism in human cells have yet to be elucidated. Therefore, these novel interactions which are required for the motility and invasiveness of human breast cancer cells, may serve as an effective approach for targeting the progression of this disease.

Rab7 regulates FA turnover, cancer cell locomotion and progression to metastasis
To understand the role of the Rab7 GTPase in cancer cell migration, we first investigated its impact on FA turnover using two human breast cancer cell lines with intrinsic invasive properties, MDA231-M2 and BT-20 cells, and their matched counterparts where Rab7 was stably knockdown by shRNA ( Figure 1A). Time-lapse confocal imaging in BT-20 cells expressing GFP-tagged paxillin revealed slower FA disassembly and persistence of FAs in Rab7-silenced cells compared to control cells ( Figure 1B, left and quantification in the right panel). A similar observation was made using the nocodazolebased assay to enable transient synchronization of FA disassembly (revealed by an enrichment of FA formation as the result of exposure to nocodazole for 2 hours followed by FA disassembly and recovery after nocodazole washout [16]. Under this condition, maximal FA disassembly was seen at 30 min post-nocodazole washout, followed by reassembly of FAs, which was obvious from 30 min to 1 h. In comparison to control cells, Rab7-silenced cells reveal slower turnover of 118Y-p-paxillin, especially at 15 min after nocodazole washout (Supplementary Figure 1).
Investigation of cell migration confirmed that in MDA231-M2 cells where Rab7 was downregulated, cell locomotion was significantly compromised compared to control cells ( Figure 1C, 1D and Supplementary Video 1). Similar results were seen in BT-20 cells. Noteworthy, reduced cell locomotion was not mediated by changes in cell proliferation as no difference in cell growth was seen between Rab7-shRNA and their matched control cells (Supplementary Figure 2).
To further confirm the correlation between these in vitro observations and cancer progression in vivo, we investigated the impact of Rab7 down-regulation on cancer cell progression to metastasis using MDA231-M2 cells expressing control shNT or Rab7 shRNA implanted orthotopically into the mammary fat pad of SCID mice. After 40 days observation, a significant inhibition in the number of lung metastases was seen with the Rab7silenced cells compared to control cells ( Figure 1E, top). Quantification confirmed that the number of lung metastases were significantly decreased in the absence of Rab7 ( Figure 1E, bottom), despite having similar primary tumors weights.

Knockdown of Rab7 promotes paxillin enrichment in cytoplasmic puncta
To further investigate the relationship between Rab7 expression and FA dynamics we examined the protein levels of various FA components in BT-20 cells under conditions where Rab7 is downregulated. As such, we observed increasing levels of phosphorylated paxillin and Src in Rab7 shRNA cells, compared to control cells (Figure 2A). Other FA-associated proteins such as FAK remained unchanged. Furthermore, we observed pronounced relocation of 118Y-p-paxillin to distinctive intracellular puncta in cells where Rab7 was downregulated, while 118Y-p-paxillin predominantly localized to adhesion sites in control cells ( Figure 2B, top and quantified in the bottom panel).
To investigate if Rab7-GTPase activity was essential for paxillin relocalization into these cytoplasmic puncta, we expressed control (GFP-C1), wild-type Rab7 (GFP-Rab7) or a Rab7-GTPase defective mutant (GFP-Rab7-T22N) [17] in control and Rab7-silenced cells ( Figure 2C). As shown in Figure 2D (solid arrows), in Rab7-deficient cells where the expression of wild type Rab7 was restored, the expression of 118Y-p-paxillin in FAs was rescued. However, expression of the dominant negative GFP-Rab7-T22N resulted in the reappearance of perinuclear 118Y-p-paxillin puncta even in control cells expressing endogenous Rab7 ( Figure 2D, left and quantification in the right panel). These findings demonstrate that interfering with Rab7 or its GTPase activity prevented the trafficking of phosphorylated paxillin.

118Y-p-paxillin accumulates in autophagolysosomes in Rab7-depleted cells
Rab7 plays an essential role in the maturation of late autophagic vacuoles [18,19]. Therefore, we investigated whether 118Y-p-paxillin was arrested in these late autophagic vacuoles. To do so, we first used chloroquine (CQ), a small molecule that accumulates in autophagic vesicles to prevent fusion of autophagosomes to lysosomes [20]. As shown in Figure 3A, exposure of cells to CQ for 24 h significantly led to the accumulation of LC3-IІ, which was similar to what we observed in cells expressing Rab7 shRNA ( Figure 3A). Moreover, both CQ and Rab7 shRNA induced LC3 puncta formation, as compared to respective controls ( Figure 3B), which indicated that both approaches cause late stage autophagy blockade. To further decipher the localization of these 118Y-p-paxillin puncta, co-staining of 118Y-p-paxillin with LAMP-1 (lysosome marker) and LC3 (autophagy marker) was performed. As shown in Figure 3C, the puncta observed upon Rab7 knockdown Oncotarget 31201 www.impactjournals.com/oncotarget or CQ treatment were indicative of an accumulation in autophagolysosomes ( Figure 3C). These findings were further supported by our density gradient centrifugation studies, which consisted of enriching various cellular compartments including autophagosomes. Although 118Y-p-paxillin accumulation in autophagosomes is clearly visible in both cell lines, increased accumulation is observed in shRab7 autophagosomes since trafficking is compromised in this condition ( Figure 3D). Furthermore, monitoring FA dynamics in live cells revealed significantly reduced FA disassembly rates and prolonged FA duration at FA site in CQ-treated cells, as compared to controls (Supplementary Figure 3). This specific localization of 118Y-p-paxillin with LC3 in autophagosomal puncta upon CQ treatment was not exclusive to BT-20 cells, as it was also observed in CQ-treated MDA-MB-231 and MCF-7 cells ( Figure 3E).

Rab7 regulates 118Y-p-paxillin turnover through the autophagy pathway rather than endosomal degradation
Considering that Rab7 can play a role in both endosomal and autophagic pathways which converge at late autophagic vacuoles, we seeked to identify which of these pathways is primarily responsible for 118Y-p-paxillin trafficking to autophagosomes. Therefore, we silenced either Rab5 (endosomal upstream regulator) or Atg12 Oncotarget 31202 www.impactjournals.com/oncotarget (autophagic upstream regulator) using siRNA ( Figure 4A) and then monitored 118Y-p-paxillin puncta staining. As shown in Figure 4B, the 118Y-p-paxillin puncta staining could be rescued in Rab7-silenced cells after knocking down Atg12, but not Rab5. In addition, knockdown of Atg12 resulted in the enhancement of focal adhesion structures both in control and Rab7-silenced cells. This 118Y-p-paxillin enhancement in Atg12-silenced cells was further confirmed by immunoblotting ( Figure 4C). To further confirm the dependence of FA turnover on the autophagic pathway, we monitored FA dynamics in Atg12knockdown cells expressing GFP-paxillin. As expected, Atg12 downregulation significantly decreased the rate of paxillin disassembly ( Figure 4D), further reinforcing the implication of the autophagic pathway in Rab7-mediated turnover of 118Y-p-paxillin.

Phosphorylation of the Y118-residue of paxillin is exclusively required for its targeting to autophagosomes and for autophagy-mediated FA turnover
To establish the importance of paxillin posttranslational tyrosine modifications necessary for Rab7mediated autophagosomal targeting, we investigated the two tyrosine residues previously established to be  (A-C) Control (shNT), Rab7-silenced (shRab7) and control treated with 20 μM chloroquine (CQ) for 24 h (shNT+CQ) BT-20 cells lysed and immunoblotted with antibodies against Rab7, LC3B and β-actin (internal control) (A), fixed and stained with anti-LC3 antibody (green) and with DAPI (blue) with 20 μm scale bar left (insert 63X) and quantification on the right panel showed percentage of cells with LC3 puncta and average # of LC3 puncta/cell (N = 15) (B). Data are presented as mean ± SEM (*p < 0.05, n = 6)) or further co-stained with anti-118Y-p-paxillin antibody (red), anti-LC3B antibody (green), anti-LAMP-1 antibody (cyan) and with DAPI (blue) with 20 μm scale bar (C). Enlargements and cross-sections of the confocal-z-planes of the boxed regions are also shown indicating association of 118Y-ppaxillin, LC3B and LAMP-1. (D) Lysed shNT and shRab7-expressing BT-20 cells were subjected to nycodenze gradient centrifugation, then analyzed by immunoblotting using anti-LAMP1, anti-118Y-p-paxillin, anti-Rab7 and anti-LC3 antibodies. LD is the loading control, A1 and A2 are autophagosome-related fractions, L is the lysosomal fraction and M is the mitochondrial fraction. (E) MDA-MB-231, BT-20 and MCF-7 cells were treated either with PBS (untreated) or CQ for 24 h, then fixed and stained with anti-118Y-p-paxillin or anti-416Y-p-Src antibodies (red) and anti-LC3 antibody (green). Scale bar, 20 μm.

Y118-p-paxillin interacts with LC3 at FA sites
In order to gain further insight into mechanisms of paxillin recruitment to autophagosomes, we co-transfected BT-20 cells with FLAG-paxillin and GFP-LC3, followed by immunoprecipitation of FLAG-paxillin. As shown in Figure 6A, we were able to co-immunoprecipitate GFP-LC3 with FLAG-paxillin. Once again, this interaction Oncotarget 31206 www.impactjournals.com/oncotarget was dependent on the phosphorylation of paxillin at Y118 since expression of the paxillin mutant Y118F, which is unable to get phosphorylated ( Figure 6B, left), failed to interact with LC3 ( Figure 6B, right). Furthermore, we co-transfected BT-20 cells with GFP-paxillin and mCherry-LC3 which allowed us to follow their intracellular localization in live cell conditions. As shown in Figure 6C, 6D) and Supplementary Video 2, transient co-localization of EGFP-paxillin-WT and mCherry-LC3 are seen at FA sites. To further confirm that this transient interaction occurs during FA disassembly, we utilized the nocodazole-based assay as previously described [22]. As expected, immunocytochemistry results demonstrated transient colocalization between 118Y-p-paxillin and LC3 at plasma membrane protrusions 10 min post nocodazole washout (Supplementary Figure 4). Likewise, expression of Y118F-paxillin resulted in significantly lower incidence of this co-localization with LC3, along with its inability to specifically co-localize at FA site ( Figure 6C, bottom). Equally important, siRNA knockdown of FAK and Src prior to co-immunoprecipitation studies had no impact on paxillin/LC3 interactions ( Figure 6E and 6F), supporting that the autophagy-mediated regulation of paxillin turnover is independent of FAK or Src kinases, both of which are implicated in the phosphorylation of paxillin [21].

c-Cbl is the main cargo receptor targeting Y118p-paxillin to autophagosomes
The E3 ubiquitin ligase activity of c-Cbl, via its LIR (light chain 3 (LC3)-interacting region) domain, has been reported to recruit Src to LC3 [23]. Additionally, c-Cbl has been reported to specifically interact with paxillin [24]. Therefore, in order to explore whether c-Cbl mediates the LC3/paxillin interaction, we used immunofluorescence staining to investigate the association. We noticed the presence and co-localization of a c-Cbl/118Y-p-paxillin/LC3 complex at focal adhesions ( Figure 7A). Next, we knocked down c-Cbl expression using siRNA ( Figure 7B, left) and then pulled down the complex. As shown in Figure 7B (right panel), c-Cbl siRNA expression considerably reduced LC3 binding to paxillin. This interaction was independent of the E3 ubiquitin ligase activity of c-Cbl, as expression of a E3 ubiquitin ligase-defective mutant of c-Cbl (HA-Cbl-C381A) still maintained an effective interaction with paxillin ( Figure 7C). Interestingly, c-Cbl siRNA also caused the accumulation of 118Y-p-paxillin in BT-20 cells, an accumulation that can be reversed by the re-expression of either WT or C381A c-Cbl ( Figure 7D). The significance of this accumulation was visualized by immunofluorescence, where enhanced 118Y-p-paxillin structures (density and number) were observed in BT-20 cells where c-Cbl was knocked down, as compared to control cells ( Figure 7E). Furthermore, monitoring of EGFP-paxillin turnover in motile cells revealed lower disassembly rates in BT-20 cells where c-Cbl is silenced, which was once again effectively rescued by re-expression of WT or C381A c-Cbl ( Figure 7F).

DISCUSSION
Cell locomotion is controlled by actin cytoskeletonassociated adhesions, represented by a network of proteins among which the scaffolding protein paxillin plays a fundamental regulatory role in the formation of both nascent and mature focal adhesions at leading edges of motile cells [25][26][27]. These functions are attributed to multiple interactions of paxillin with FA partners, primarily involving its zinc-finger motifs and are tightly regulated by paxillin posttranslational modifications, including phosphorylations at tyrosine residues 31 and 118 [28][29][30][31]. The role of these phosphorylation sites in the regulation of FA dynamics is still debated since both phosphorylated and non-phosphorylated paxillin have been proposed to regulate lamellipodial protrusions [32,33], as well as formation [34,35] and turnover of adhesion formation [21].
Our study identified the striking impact of Rab7 (implicated in transport from early to late endosomes of late endocytic structures/lysosomes) in the regulation of paxillin tyrosine-118 phosphorylation turnover via the autophagy pathway, an evolutionary conserved catabolic process which delivers cytoplasmic cargo to lysosomes via double membrane vesicles called autophagosomes [36] but not through proteasomal degradation [37]. Knockdown of Rab7 but not Rab5 (a regulator of early endosome biogenesis) or Rab11 (involved in perinuclear recycling of endocytosed proteins) (data not shown) prevented 118Y-p-paxillin recruitment and accumulation in the autophagosomes ( Figure 4B). As well, we confirmed that the Rab7 mutant lacking GTPase activity failed to rescue118Y-p-Pax recruitment to autophagosomes, highlighting the importance of Rab 7 GTPase activity ( Figure 2D), which requires a switch between the GDPbound off-state and a GTP-bound on-state, necessary for paxillin recognition and recuitment to autophagosomes.
The critical function of Rab7 to recruit autophagy effector FA proteins is of fundamental importance given that autophagy, an evolutionary conserved catabolic process which delivers cytoplasmic cargo to lysosomes via autophagosomes [36] , is implicated in the regulation of cell migration [38,39]. In our study, downregulation of autophagy using both genetic and pharmacological approaches resulted in selective accumulation of 118Y-ppaxillin but not Y31-p-paxillin in autophagosomes ( Figure 5A). These results were corroborated by imaging studies confirming intracellular colocalization of 118Y-ppaxillin but not Y31 with LC3 at cell protrusions during FA disassembly ( Figure 5C, 5D). In contrast to a previous study showing that mutation of both Y31 and Y118 to non-phosphorylatable amino acids impairs the disassembly of adhesions at FA sites of the leading edge www.impactjournals.com/oncotarget  Oncotarget 31209 www.impactjournals.com/oncotarget of migrating cells [21,40], our study revealed a distinct function of Y118, which unlike Y31, can selectively target paxillin to autophagosomes for degradation, triggering FA disassembly.
Paxillin has a proline-rich motif that binds to the Src SH3 domain and Src-FAK kinases, which mediate tyrosine phosphorylation of paxillin [41]. In turn, paxillin interacts with the FAT domain of FAK believed to direct FAK to FA sites. Furthermore, it has been shown that it is the phosphorylated form of paxillin that predominantly recruits FAK into the adhesion sites and that paxillin-induced adhesion turnover occurs in an FAK-dependent manner [40]. To rule out the possibility that paxillin recruitment to autophagosomes is a result of a complex involving Src or FAK, we demonstrated that Src or FAK knockdown had no impact on Y118-p-paxillin recruitment to autophagosomes or interaction with LC3B ( Figure 6E, 6F), suggesting that recruitment to autophagosomes occurs in a Src-and FAKindependent manner. In fact, as shown in Supplementary Figure 5, neither FAK, 397Y-p-FAK nor 861Y-p-FAK accumulated in puncta in Rab7-deficient cells, compared to matched control cells.
During selective autophagy, recognition of protein targets involves direct interaction between the substrate and LC3 through LIR motifs, which include cargo receptor proteins such as c-Cbl, NBR1 and p62 [42]. Interestingly, while writing this manuscript a recent study reported that autophagy regulates FA turnover via the cargo protein NBR1 [43]. In our study, using siRNA targeting NBR1 prior to pull down revealed that although NBR1 regulates EGFP-LC3/FLAG-paxillin interaction and colocalizes with 118Y-p-paxillin at FA site, the efficiency of Y118-phospho-paxillin localization to autophagosomes is weaker compared to when manipulating c-Cbl (Supplementary Figure 6). Therefore, we conclude that c-Cbl is a major cargo receptor but not the ubiquitinbinding scaffolding protein p62, which also interacts with LC3 during autophagy. c-Cbl has ubiquitin ligase activity and one study reported that inhibition of c-Cbl ubiquitin ligase activity prevented paxillin interaction and degradation in myocytes [44]. However, in our study we demonstrate that c-Cbl ligase activity had no impact on paxillin/c-Cbl interaction neither on 118Y-ppaxillin expression and turnover ( Figure 7D, 7F). In support of our findings, autophagic targeting of active Src is reported to be mediated by c-Cbl independently of c-Cbl E3 ligase activity [23]. Nevertheless, we cannot rule out a contribution of additional posttranslational modifications of paxillin such as serine phosphorylation [37,45] and ubiquitination [46], both deserving further investigations. Moreover, in agreement with previously published studies, paxillin phosphorylation at Y118 may result in a conformational change in paxillin creating a Crk binding site [30], which in turn possesses a cbl binding domain [47], thereby aiding in paxillin recruitment and detachment from FA complexes. Consistent with recent observations described by Sharifi et al. [15], this is in sharp contrast to what we observed in mouse derived cell lines, where phosphorylation at Y118-p-paxillin was not required for autophagosomal regulation of paxillin and/ or its interaction with LC3 in the mouse mammary tumor derived cell line 4T1, or in FAK −/− or SYF −/− mouse embryonic fibroblasts (data not shown), further supporting a tissue-type specific mechanism for this turnover.
In summary, our data provides new mechanistic insights into the role of Rab7 in targeting Y118 sitespecific phosphorylation of paxillin to the autophagy pathway to promote focal adhesion turnover (summarized in Supplementary Figure 7) in human breast cancer cells, a process essential for cell locomotion. Noticeable, many of the protein networks involved in these pathways have been reported to be upregulated in many human cancers and can predict invasiveness, these include Rab7 [12] in agreement with another study in melanoma [13], paxillin [48] and Atg proteins [49,50]. Therefore, molecular mechanisms described herein further enlighten the molecular basis involved in these associations.

Immunoblotting assay
Sub-confluent cells were washed with PBS, lysed in RIPA buffer (50 mM Tris-HCl at pH7.5, 150 mM sodium chloride, 1% tritonX-100, 0.1% SDS, 2mM EDTA and 25mM sodium fluoride) supplemented with 1mM PMSF and protease inhibitor cocktail (Roche) for 10 min on ice and centrifuged (13,000 rpm at 4°C for 20 min) to separate cell lysate. Cell lysate (50 μg protein, as measured by the Bradford protein assay) was then added with SDS sample buffer (Tris at pH 6.8, 20% glycerol, 5% SDS, bromophenol blue and β-mercaptoethanol) and boiled for 5 min. Samples were then resolved through 13% SDS-PAGE gels, transferred to nitrocellulose membrane, blotted with primary antibodies at different dilution in cold room overnight, and then amplified with horseradish peroxidase-conjugated secondary antibodies for 1h in room temperature and enhanced by chemiluminescence detection system.

Immunofluorescence microscopy
Cells were grown on 18-mm cover glass coated with 5 μg/ml fibronectin for 24 h in 4°C, placed in 6-well culture plate for 24 h, rinsed in PBS, fixed with 4% paraformaldehyde/PBS for 10 min, washed twice in PBS with 0.2% TritonX-100, blocked in PBS with 0.2% Triton X-100 containing 1% BSA (Bioshop) and incubated with primary antibodies overnight in 4°C (all in 1/100 diluted by blocking solution). The cells were washed with PBS containing 0.2% TritonX-100 and subsequently incubated with Alexa Fluor 488-labelled, Alexa Fluor 594-labelled, or Alexa Fluor 647-labelled secondary antibodies (all were used at 1/500 dilution in blocking solution) for 1 hr in room temperature. The nuclei were stained with DAPI (0.1 μg/ml) for 5 min before mounted with aqueous mounting medium. Cells were imaged using WaveFX spinning disk confocal microscope system (Quorum Technologies INC.). Images shown are representative of three independent experiments

Live cell imaging microscopy
For the monitoring of FA turnover, cells were transfected with EGFP-paxillin and plated in RPMI 1640 serum-free medium on a multiwell chambered coverglass (LabTek) prepared by coating overnight in 4°C with 5 μg/ml fibronectin. The coverglass was directly placed on a heated stage with 5% CO 2 and stimulated with 20 ng/ml epidermal growth factor. Fluorescent images were captured every 5 min for 2 h using a heated 64 X NA 1.40 objective at WaveFX spinning disk confocal microscope system (Quorum Technologies INC.). Fluorescence intensities of individual adhesions were measured over time by Volocity imaging software (PerkinElmer) and quantified according to the previously described protocol (Franco et. al 2004b;Webb et al., 2004). Measurements were made on at least 20 individual adhesions in four separated cells for both Rab7 silenced and control groups. For investigating colozalization, cells were transfected with both GFP-paxillin and mcherry-LC3, then plated in RPMI 1640 medium with 10% FBS on a multiwell chambered coverglass (LabTek). Fluorescent images were captured every 15 sec for 2 h using a heated 63 X NA 1.40 objective using a WaveFX spinning disk confocal microscope system.

Nocodazole-based FA disassembly assay
FA disassembly and reformation, cells were grown on glass-coverslips coated with 5 μg/ml fibronectin for 24 h in 4°C, serum starved for 4 h in RPMI1640 serum free medium and then treated with 2μM nocodazole in serumfree medium for 2 h to allow a complete depolymerization of microtubules. Nocodazole was washed-out with serum-free medium and cells were further incubated at 37°C to allow a reformation of FA structure. Subsequently, cells were fixed and processed for immunofluorescence staining at different time intervals. Cells were stained with Rab7, 118Y-ppaxillin and LC3B. Focal adhesion dynamics was analyzed as previously described [22] using Image J software.

Live cell locomotion assay
Cells were seeded at very low density in serum-free medium on a multiwell chambered coverglass (LabTek) prepared by coating overnight in 4°C with 5 μg/ml fibronectin. The coverglass was placed on a heated stage with 5% CO 2 and stimulated with 20 ng/ml epidermal growth factor before monitoring the cell movement using a heated 40 X NA 1.40 objective at WaveFX spinning disk confocal microscope system (Quorum Technologies INC.). Single cell speed was determined by manually tracing the cell periphery every 20 min for 2 h by using Volocity imaging software (PerkinElmer).

In vitro statistical analysis
All experiments counting 118Y-p-paxillin and LC3 puncta were performed in triplicate with similar results. Quantitative data in single cell migration experiments and puncta quantification were performed in 15 separate cells for both Rab7 silenced and control cells. For FA disassembly and FA life time quantification, measurements were made on at least 40 focal adhesions in 10 separate Rab7 silenced and control cells. All data are presented as mean ± SEM using Prism (GraphPad Software). Statistical significance was analyzed using unpaired two-tailed Student's t test. Data were deemed to be statistically significant if P < 0.05. Error bars indicate SEM unless otherwise indicated. www.impactjournals.com/oncotarget

In vivo xenograft model of breast cancer metastasis
All experiments were carried out according to protocol number 4101 of the McGill University Animal Care Committee. MDA231-M2 cells (1 × 10 6 cells/ each mouse) expressing scramble shRNA and their matched cells stably expressing Rab7 shRNA were transplanted into the mammary fat pad of SCID mice. Tumor size was measured using a caliper, and tumor volume was calculated as π/6 (length × width 2 ). Primary tumors were excised once they had reached a mean size of 0.8 cm 3 . The wound was closed with a single layer of surgical clips. Mice were sacrificed 40 days after transplanting. The lungs were fixed in 10% Bouin's fixative and lung metastatic nodules on the surface were counted using a stereomicroscope (Optimax; Leica). Statistical comparison between groups was performed using a computer-based statistical package from Statistical Product and Service Solution (Chicago, IL) as we described earlier [52].