Autophagy-Dependent Production of Secreted Factors Facilitates Oncogenic RAS-Driven Invasion

The RAS proteins are members of a family of small GTPases critical in mediating cellular responses following activation by upstream extracellular signals, such as growth factors. Oncogenic mutations in RAS, which result in constitutive activation, are found in approximately 30% of human cancers; they are highly prevalent in several carcinomas, including lung, pancreas, and colon (1, 2). Notably, oncogenic RAS drives diverse cellular programs—proliferation, cell survival, migration, invasion, and alterations in differentiation—that support tumor initiation and progression. Such mutations present a formidable therapeutic obstacle, because patients harboring mutant KRAS are refractory to most available systemic therapies and exhibit extremely poor survival (2). Hence, identifying new processes to target cancer cells with hyperactive RAS remains a question of immense clinical significance. One such pathway may be macroautophagy (autophagy), a tightly controlled lysosomal degradation process that promotes cell survival during nutrient starvation and stress. Recent evidence indicates that basal autophagy levels are enhanced upon oncogenic RAS activation and support RAS-driven transformation and tumorigenesis (3–7).

The tumor-promoting functions of autophagy are largely ascribed to its importance as a survival pathway in response to diverse environmental stresses (8, 9). For example, enhanced autophagy is observed in poorly perfused, hypoxic tumor regions, and loss of autophagy is associated with increased necrosis (10). Autophagy also promotes tumor cell survival in response to various cytotoxic and targeted chemotherapies (11). Importantly, studies of oncogenic RAS transformation have revealed that the protumor effects of autophagy are not limited to increased survival of cancer cells under duress; rather, autophagy contributes to the metabolic fitness of the entire tumor population (3–6). Because strong oncogenic insults, such as RAS activation, are marked by profound metabolic alterations that drive both energy production and biosynthetic capacity in rapidly proliferating cells, it has been hypothesized that autophagy maintains key metabolic pathways in RAS-transformed cells. In support, a growing body of work has unveiled a requirement for autophagy in driving proliferation as well as sustaining multiple core metabolic functions in RAS-transformed cells (3–7). These results are not unique to oncogenic RAS activation, as deletion of RB1CC1/FIP200, a mediator of autophagosome initiation, inhibits polyoma middle T–driven mammary cancer, due to reduced proliferation and glucose metabolism (12).

In addition to its effects on proliferation and metabolism, oncogenic RAS drives diverse aggressive cellular behaviors that support tumor progression and metastasis; importantly, RAS-transformed epithelial cells exhibit highly invasive behavior associated with an epithelial-to-mesenchymal transition (EMT; ref. 13). Here, in epithelial cells transformed with oncogenic RAS, we demonstrate that autophagy facilitates extracellular matrix (ECM) invasion, tumor cell motility, and pulmonary metastasis in vivo. Using a three-dimensional (3D) culture system, we uncover that autophagy inhibition restricts RAS-driven cell invasion and restores several aspects of normal epithelial architecture, including the polarized deposition of basement membrane and cell–cell junctional integrity. Furthermore, autophagy is required for the production of multiple secreted factors in RAS-transformed cells, including interleukin-6 (IL-6), matrix metalloproteinase 2 (MMP2), and WNT5A, which altogether facilitate cancer cell invasion.

To elucidate how autophagy affects the cellular behavior of RAS-transformed epithelial cells, we used the MCF10A 3D epithelial culture system to interrogate how autophagy affects the growth and morphogenesis of cells expressing oncogenic RAS (14). We generated stable pools of MCF10A human mammary epithelial cells expressing a control vector (BABE) or an oncogenic form of HRAS (HRAS^V12) that enhances basal autophagy and elicits robust anchorage-independent transformation (3). When cultured on laminin-rich ECM, control MCF10A cells formed hollow, spherical acini (Supplementary Fig. S1A; ref. 15). In contrast, HRAS^V12-transformed cells produced grossly aberrant structures notable for extensive protrusions that invaded the surrounding ECM. Individual HRAS^V12 structures formed these invasive protrusions in as early as 3 to 5 days, ultimately producing disorganized networks of cells intermingled with large cell clusters after 8 days in 3D culture (Fig. 1A and B, left). The 3D morphology we observed using HRAS^V12 MCF10A cells resembles that reported for mouse mammary cells expressing oncogenic RAS and grown in a 3D collagen matrix (16).

To inhibit autophagy in this experimental system, we stably expressed unique short hairpin RNAs (shRNA) against two autophagy genes (ATG)—ATG7 (shATG7-1 and shATG7-2) or ATG12 (shATG12)—in MCF10A cells expressing HRAS^V12. ATG7 or ATG12 knockdown decreased target protein levels, reduced basal- and starvation-induced autophagy in Hank's Buffered Saline Solution (HBSS), and increased protein levels of the autophagy substrate p62/SQSTM1 (Supplementary Fig. S1B–S1E). In 3D culture, the invasive protrusions observed with oncogenic RAS activation were profoundly attenuated in ATG-deficient cells. Instead, HRAS^V12 shATG (shRNAs against autophagy genes) structures were spherical in morphology, similar to nontransformed BABE controls (Fig. 1A and B). Decreased invasive protrusions following autophagy inhibition were also observed upon stable ATG3 knockdown (shATG3), and upon treatment with chloroquine or bafilomycin A, two lysosomal inhibitors that block the late steps of autophagy (Supplementary Fig. S1F). Importantly, ATG knockdown in HRAS^V12 cells did not affect RAS expression– or activation–associated phosphorylation of the major downstream effector MAPK–ERK (Supplementary Fig. S1G). Thus, the reduction in 3D invasive protrusions following ATG knockdown is not due to decreased expression or activity of oncogenic RAS.

The disruption of basement membrane integrity is a hallmark of carcinoma invasion in vivo (14). To corroborate whether the protrusions we observed in HRAS^V12-transformed 3D cultures represented invasive behavior, we first evaluated basement membrane integrity by examining the expression and localization of the basement membrane protein LAMA5 (laminin 5) in HRAS^V12-derived acini. Consistent with previous reports, control (BABE) nontransformed MCF10A acini displayed polarized deposition of LAMA5 onto the basal surface (Fig. 2A, left; ref. 15). In contrast, the expression of HRAS^V12 resulted in cytosolic accumulation of LAMA5, with no evidence of polarized deposition at the cell–ECM interface. Notably, this aberrant cytosolic staining pattern was especially prominent in the protrusions of HRAS^V12 cultures. Correlating with the decreased formation of invasive protrusions, ATG knockdown restored polarized LAMA5 secretion; based on this marker, most individual structures in ATG-deficient HRAS^V12 cultures were encompassed by an intact basement membrane (Fig. 2A). Hence, in addition to restricting the formation of invasive protrusions, autophagy inhibition restored polarized basement membrane secretion typically absent in HRAS^V12 shCNT (nontargeting control shRNA) structures.

To extend these results, we evaluated ECM proteolytic activity in control and autophagy-deficient HRAS^V12 cultures by assessing fluorescence emanating from the proteolytic cleavage of dye-quenched collagen IV (DQ-COL4). In control (BABE) nontransformed acini, we observed a faint ring of fluorescence surrounding each structure, corresponding to COL4 degradation due to the normal outgrowth of acini during 3D morphogenesis. On the other hand, HRAS^V12 shCNT-expressing structures exhibited high levels of fluorescence that extended well beyond the immediate vicinity of individual structures (Fig. 2B). Notably, streaks of fluorescence connecting adjacent structures were frequently observed in HRAS^V12 shCNT cultures (Fig. 2B), which resembled the networks of invasive protrusions (Fig. 1B). In contrast, HRAS^V12 shATG-derived structures exhibited a ring-like COL4 degradation pattern that was restricted to the cell–ECM interface, similar to that observed in nontransformed controls (Fig. 2B). Thus, the absence of morphologic protrusions in ATG-deficient HRAS^V12 cultures was associated with the restoration of basement membrane integrity and reduced ECM proteolytic activity. Together, these findings corroborate that autophagy supports RAS-driven invasion in 3D culture.

We next evaluated the impact of autophagy inhibition on oncogenic RAS-driven proliferation and cell survival. During normal MCF10A acinar morphogenesis, autophagy inhibition results in the enhanced apoptosis of cells occupying the luminal space (17). To test whether autophagy deficiency similarly affected apoptosis in HRAS^V12 structures, we immunostained structures with an antibody against cleaved CASP3 (caspase-3). In contrast to the robust luminal apoptosis observed in control acini (BABE), only isolated cleaved CASP3-positive cells were observed in HRAS^V12 shCNT structures, consistent with the ability of oncogenic RAS to promote cell survival in 3D culture (Fig. 3A). Upon enumerating cleaved CASP3-positive cells from these 3D cultures, we found that ATG knockdown did not significantly affect apoptosis in comparison with shCNT cultures (Fig. 3A). To assess whether autophagy inhibition potentially affected nonapoptotic death processes, we also stained day 8 3D cultures with ethidium bromide (EtBr), an intravital dye that is incorporated into all dying cells. Although acini derived from nontransformed (BABE) cells displayed high levels of EtBr staining corresponding to luminal cell death (Fig. 3B), HRAS^V12 structures displayed only occasional EtBr cells scattered throughout the structures. Although ATG knockdown in HRAS^V12 cultures resulted in spherical structures that lacked invasive protrusions, we did not observe any increase in EtBr staining in these cultures (Fig. 3B). Thus, in contrast to normal and oncogenic PIK3CA MCF10A acinar morphogenesis, autophagy inhibition does not promote apoptosis in RAS-transformed 3D structures (17, 18).

To evaluate the effects of autophagy inhibition on the proliferative capacity of HRAS^V12 structures, we immunostained cultures with the proliferation marker Ki67 on day 8, a time point at which normal MCF10A acini exhibit reduced proliferation (19). As expected, low levels of Ki67-positive cells were observed in BABE structures (Fig. 3C, left). However, both control and autophagy-deficient HRAS^V12 structures displayed high levels of Ki67-positive cells (Fig. 3C). Overall, these results indicate that although autophagy deficiency potently restricts HRAS^V12-driven invasion, it does not universally suppress the diverse oncogenic effects of HRAS^V12 in 3D culture, including the ability of activated RAS to inhibit apoptosis and sustain proliferation.

Because defects in invasive capacity are often associated with diminished cell motility, we next measured cell migration in autophagy-competent and -deficient epithelial cells. Upon ATG depletion, HRAS^V12 MCF10A cells demonstrated an approximately 30% reduction in migratory capacity in a monolayer wound-healing assay of cell migration (Fig. 4A). Similar results were obtained using a Transwell migration assay, which demonstrated a significant decrease in migration of ATG knockdown cells (Fig. 4B). We further corroborated these results using MDA-MB-231 cells, a highly migratory, KRAS-mutant breast cancer cell line. siRNA-mediated knockdown of either ATG7 or ATG12 in MDA-MB-231 cells resulted in reduced LC3-II formation (Supplementary Fig. S1H) as well as decreased wound closure (Fig. 4C, left). A similar decrease in MDA-MB-231 migration was also observed in the presence of the lysosomal inhibitor bafilomycin A (Fig. 4C, right). Therefore, in addition to supporting invasion of HRAS^V12 MCF10A cells in 3D culture, autophagy facilitates the migration of cells expressing oncogenic RAS in monolayer culture. Finally, we used an experimental metastasis assay to evaluate whether the effects of autophagy inhibition on invasion and migration correlated with changes in metastatic capacity in vivo; in support, the ability of HRAS^V12 MCF10A cells to produce pulmonary metastases was reduced upon ATG knockdown (Fig. 4D).

Constitutive RAS activation alters epithelial differentiation by driving an EMT (20, 21), a process associated with increased invasive and migratory capacity in vitro and with metastatic capacity in vivo (13). Therefore, we evaluated how autophagy inhibition affects protein expression changes associated with RAS-induced EMT. We isolated BABE-, HRAS^V12 shCNT-, and HRAS^V12 shATG-expressing cells from day 8 3D cultures and determined the protein expression of a panel of EMT-associated genes by immunoblotting. In comparison with nontransformed BABE acini, HRAS^V12 shCNT structures displayed decreased keratin 14 (KRT14), an epithelial marker, and a corresponding increase in the mesenchymal protein vimentin (VIM; Supplementary Fig. S2A). ATG knockdown reversed these HRAS^V12-driven changes in differentiation, resulting in an increase in KRT14 protein levels and a corresponding decrease in VIM levels compared with HRAS^V12 shCNT cells isolated from 3D culture (Supplementary Fig. S2A). However, autophagy inhibition had minimal effects on other EMT markers that were altered by oncogenic RAS expression. Only a slight increase in E-cadherin (CDH1) was observed in shATG cells, decreased fibronectin (FN1) was only observed in shATG7-1–expressing cells, and N-cadherin (CDH2) levels were unchanged following ATG knockdown (Supplementary Fig. S2A).

During EMT, cells commonly lose the ability to form cell–cell junctions (22). Therefore, we analyzed the effects of autophagy inhibition on cell–cell junctional integrity in HRAS^V12 3D structures by immunostaining for β-catenin (CTNNB1). Normal MCF10A acini (BABE) displayed strong β-catenin staining at cell–cell contacts, indicating intact adherens junctions, whereas the expression of HRAS^V12 resulted in a near-complete loss of β-catenin junctional staining; in these cultures, only isolated focal areas of junctional β-catenin staining were observed (Supplementary Fig. S2B). Upon ATG knockdown in HRAS^V12 structures, both the expression and junctional localization of β-catenin were significantly restored (Supplementary Fig. S2B). On the basis of these results, we conclude that autophagy inhibition modulates certain aspects of mesenchymal differentiation in RAS-transformed cells in 3D culture, most notably the suppression of VIM, as well as the restoration of KRT14 expression and epithelial cell–cell contacts. Nonetheless, autophagy deficiency does not broadly suppress RAS-driven EMT.

Cell migration and invasion involves the secretion of multiple factors that cooperate to promote motility and to degrade the surrounding ECM (23, 24). To ascertain whether defects in RAS-driven invasion observed following autophagy suppression were the result of decreased production of proinvasive factors, we performed a coculture assay in which HRAS^V12 shATG7-1 cells (coexpressing GFP for tracking purposes) were combined with HRAS^V12 shCNT cells at a ratio of 3:1, respectively. Although HRAS^V12 shATG7-1–GFP cells cultured alone grew as spherical structures (Fig. 5A, left), upon coculture with HRAS^V12 shCNT cells, HRAS^V12 shATG7-1–GFP structures became dispersed and formed invasive protrusions (Fig. 5A, right). Hence, we hypothesized that factors from neighboring HRAS^V12 shCNT cells are sufficient to rescue in trans the invasion defect in HRAS^V12 shATG7-1 cells. To further test this prediction, we grew HRAS^V12 shATG cells in 3D culture for 3 days and subsequently treated these structures with conditioned media produced from either BABE or HRAS^V12 shCNT cultures. HRAS^V12 shATG structures remained as compact spheres following treatment with BABE conditioned media (Fig. 5B and Supplementary Fig. S3A). In contrast, conditioned medium from HRAS^V12 shCNT cultures elicited invasive protrusions at 24 hours following treatment, which became fully evident by 72 hours (Fig. 5B and Supplementary Fig. S3A); notably, conditioned medium addition did not induce invasion in nontransformed BABE acini (Supplementary Fig. S3B). Furthermore, basement membrane integrity was lost in HRAS^V12 shATG cells treated with HRAS^V12 conditioned medium (Fig. 5C). These findings demonstrate that autophagy inhibition in HRAS^V12 cells inhibits the production of secreted factors required for RAS-driven invasion in 3D culture.

During RAS-induced senescence, ATG depletion inhibits IL-6 production following acute oncogenic RAS activation in IMR90 fibroblasts, indicating that autophagy supports the production of IL-6 in response to oncogenic RAS activation (25). Because IL-6 has been demonstrated to support RAS-driven tumorigenesis, promote migration, and invasion, and also drive EMT (26–28), we tested whether IL-6 levels were altered in HRAS^V12 shATG 3D cultures. Analysis of IL-6 in conditioned media collected from 3D cultures by ELISA indicated a significant reduction in secreted IL-6 levels in HRAS^V12 shATG-expressing cultures compared with HRAS^V12 shCNT cultures (Fig. 6A). Furthermore, this decrease in secreted IL-6 was not the result of reduced IL6 gene expression; in fact, quantitative PCR (qPCR) analysis revealed that IL6 transcript levels in HRAS^V12 shATG cells were increased, rather than decreased, in comparison with HRAS^V12 shCNT cells (Fig. 6B). Notably, studies of RAS-induced senescence similarly demonstrated that autophagy-deficient cells exhibit reduced IL-6 protein levels due to impaired translation, rather than transcription (25, 29). In contrast, we uncovered that ATG depletion did not attenuate IL-6 protein levels in RAS-transformed cells grown in 3D culture (Fig. 6C). These results suggest that autophagy facilitates IL-6 secretion during HRAS^V12 3D morphogenesis.

To ascertain the functional significance of these results, we interrogated whether IL-6 was necessary for HRAS^V12-driven invasion in 3D culture. First, we treated HRAS^V12 shATG structures with HRAS^V12 shCNT conditioned media in the presence versus absence of an IL-6 function–blocking antibody. The addition of IL-6 function–blocking antibody attenuated the ability of HRAS^V12 shCNT conditioned media to promote invasive protrusions in HRAS^V12 shATG cultures, whereas an immunoglobulin G (IgG) isotype control had no effect (Fig. 6D). In parallel, we tested how exogenous recombinant human IL-6 (rhIL-6) treatment affected HRAS^V12 shATG cells during 3D morphogenesis. rhIL-6 addition did not affect nontransformed BABE acini (Supplementary Fig. S3C) but partly restored invasion in HRAS^V12 shATG cultures, resulting in large globular structures, increased invasive protrusions, and loss of basement membrane integrity (Fig. 6E and Supplementary S3D–S3E). Also, rhIL-6 addition partially reversed the effects of autophagy inhibition on KRT14 and VIM expression in HRAS^V12 shATG7 cells (Supplementary Fig. S3F). Hence, our results suggest that autophagy promotes efficient IL-6 secretion by HRAS^V12 cells in 3D culture, which is necessary for invasion.

In addition to identifying a defect in IL-6 production following ATG knockdown, we performed a qPCR array to measure the expression levels of genes involved in EMT and invasion, and identified WNT5A and MMP2 as two candidate factors whose expression was upregulated in HRAS^V12 cells relative to BABE cells but potently suppressed upon autophagy inhibition. qPCR analysis of cells collected from 3D cultures confirmed a 2-fold decrease in MMP2 and WNT5A expression in HRAS^V12 shATG cells compared with HRAS^V12 shCNT (Fig. 7A and B). Notably, we also evaluated the effects of rhIL-6 treatment on MMP2 and WNT5A expression in shATG7-1 cultures and found that this was not sufficient to rescue expression, indicating that regulation of these factors was independent of IL-6 (Supplementary Fig. S3G).

Because these secreted factors have been implicated in cell migration and invasion, we further evaluated whether their decreased expression following ATG knockdown also contributed to the reduced invasive potential of HRAS^V12 shATG cells. First, we used gelatin zymography to assess MMP2 activity in conditioned media from 3D cultures. MMP2 activity was enhanced in HRAS^V12 cells compared with nontransformed (BABE) controls, and upon ATG knockdown in HRAS^V12 cells, this activity was reduced (Fig. 7C). The increase in MMP2 expression and secretion following constitutive RAS activation was necessary for RAS-driven invasion, as addition of an MMP2 inhibitor, Arp-100, was sufficient to inhibit the formation of invasive protrusions in HRAS^V12 3D cultures (Fig. 7D). Furthermore, the decrease in WNT5A expression correlated with a decrease in WNT5A protein levels in HRAS^V12 shATG cells isolated from 3D culture (Fig. 7E). Moreover, the addition of recombinant WNT5A to HRAS^V12 shATG7-1 3D cultures promoted the dissociation of cells within the structures and enhanced the formation of invasive protrusions (Fig. 7F). Thus, in addition to IL-6, autophagy facilitates the production of multiple secreted promigratory and invasive factors that support RAS-driven invasion in 3D culture.

Our results delineate a previously unrecognized function for autophagy in facilitating oncogenic RAS-driven invasion and migration. Using a 3D culture system, we demonstrate that suppression of autophagy in HRAS^V12 MCF10A cells restricts the formation of invasive protrusions, restores basement membrane integrity, and attenuates ECM proteolysis. In addition, autophagy inhibition diminishes cell migration in vitro and pulmonary metastasis in vivo. Upon treatment with conditioned media produced from autophagy-competent HRAS^V12 cells, invasion is completely restored in autophagy-deficient HRAS^V12 cultures, indicating that autophagy mediates the production of secreted factors that drive invasion in oncogenic cells. In further support, we uncover that autophagy inhibition elicits the coordinate reduction of multiple molecules favoring invasion. Overall, these findings expand our understanding of how autophagy supports cancer progression.

Although autophagy inhibition suppresses invasion in 3D culture, it does not ubiquitously revert oncogenic RAS-driven changes in cell behavior. Indeed, MAPK activation remains unaltered following autophagy inhibition in this 3D culture model, and, moreover, the oncogenic activation of RAS continues to disrupt fundamental aspects of 3D morphogenesis in autophagy-deficient cells. First, autophagy inhibition does not alter the ability of HRAS^V12 to suppress apoptosis in 3D culture. Moreover, autophagy inhibition does not suppress proliferation in HRAS^V12 3D cultures; rather, the spherical structures from HRAS^V12 ATG knockdown cells remain highly proliferative over extended periods. Remarkably, both others and we have shown that ATG depletion reduces soft-agar growth and attenuates the proliferation of RAS-transformed cells grown in monolayer (3–5, 30); hence, the absence of proliferative suppression in this 3D culture model may be context dependent. These results also differ from those obtained in KRAS-mutant mouse cancer models in which genetic ATG deletion impairs proliferation and, in certain cases, enhances apoptosis (4, 6, 7). Certain reasons may explain these differences. First, we have only reduced ATGs using RNA interference (RNAi), rather than genetically eliminated these proteins. Second, the experiments here are of significantly shorter duration in comparison with autophagy-deficient KRAS-mutant tumor growth in vivo.

Although previous studies have demonstrated that autophagy supports the invasion of glioblastoma cells, the mechanistic underpinnings remain unclear (31, 32). Cell invasion requires the production and secretion of factors that stimulate migration and degrade the surrounding ECM (24). Upon treatment of autophagy-depleted HRAS^V12 cells with conditioned media produced from their autophagy-competent counterparts, the ability to form invasive protrusions is completely restored, suggesting that autophagy is required for the efficient production of secreted factors that promote invasion and migration of HRAS^V12 cells. Notably, conditioned media treatment does not promote invasion in nontransformed BABE cells, indicating that oncogenic RAS pathway activation is still required for invasion.

Importantly, we identify IL-6 as one critical factor whose secretion is ATG dependent; our results substantiate that this proinvasive cytokine is necessary to restore invasion in autophagy-deficient HRAS^V12 cells. They also point to a specific role for autophagy in facilitating IL-6 secretion; upon ATG knockdown, RAS-transformed cells fail to secrete IL-6 into the conditioned media, yet both IL-6 transcription and translation remain intact. These results differ from recent studies of oncogenic RAS-mediated senescence, in which reduced IL-6 secretion in autophagy-deficient cells is proposed to be secondary to decreased protein synthesis (25, 29).

Although it has been traditionally viewed as an autodigestive process, growing evidence suggests new roles for autophagy in both conventional and unconventional secretion (33). Indeed, a genetic role for ATGs has been implicated in (i) unconventional secretion of proteins lacking N-terminal endoplasmic reticulum signal sequences (34–37), (ii) efficient egress of secretory lysosomes (38, 39), and (iii) conventional secretion of growth factors (40, 41). Further dissection of how autophagy directs the secretion of IL-6 and other factors during RAS transformation remains an important topic for future study. Remarkably, IL-6 re-addition only partially restores invasion and mesenchymal differentiation in HRAS^V12 autophagy-deficient cultures, indicating that other factors promote invasion. In support of this possibility, these cells exhibit reduced levels of other proinvasive molecules, including WNT5A and MMP2. In contrast to reduced IL-6 secretion, which is likely a proximal event following ATG knockdown, these changes in WNT5A and MMP2 result from decreased gene expression, indicating that autophagy inhibition produces broader transcriptional changes contributing to reduced invasion by HRAS^V12 cells.

Recently, the deletion of RB1CC1/FIP200, a gene mediating autophagosome initiation, was demonstrated to reduce lung metastases in the MMTV-PyMT breast cancer model. However, as RB1CC1 deletion profoundly restricted primary tumor growth, it was unclear whether decreased metastasis was secondary to reduced primary tumor burden (12). In addition, although liver-specific deletion of ATG7 or ATG5 initiates the development of benign adenomas, these tumors are unable to progress to adenocarcinomas, suggesting that autophagy is required for advanced tumor progression (42, 43). Here, in epithelial cells transformed with oncogenic RAS, we demonstrate that defective autophagy results in decreased invasion and migration, which correlates with the reduced ability to metastasize in vivo. Although our results do not rule out potentially important functions for autophagy in disseminated cell survival or outgrowth at foreign tissue sites, they delineate new roles for autophagy in the control of secretion during carcinoma progression.

MCF10A cells were obtained from the American Type Culture Collection (ATCC) and cultured as previously described (44). MDA-MB-231 cells were obtained from the ATCC and cultured in Dulbecco's Modified Eagle Medium (DMEM) supplemented with 10% FBS, penicillin, and streptomycin. Cell lines were passaged for less than 6 months following resuscitation and were not authenticated.

MCF10A overlay 3D culture was performed as previously described (44). As indicated, the following reagents were added to cultures: 500 ng/mL WNT5A (R&D Systems), 200 ng/mL IL-6 (PeproTech), 25 μg/mL anti–IL-6 function–blocking antibody (R&D Systems), 25 μg/mL IgG control antibody (BD Biosciences), 25 μmol/L Arp-100 (Santa Cruz Biotechnology), 5 μmol/L chloroquine diphosphate salt (Sigma), and 5 nmol/L bafilomycin A (Sigma). For the 3D ECM degradation assay, human DQ-COL4 (Invitrogen) was mixed with Matrigel to a final concentration of 25 μg/mL before plating. To collect cells for immunoblotting and RNA isolation, cultures were incubated with 0.25% Trypsin/EDTA at 37°C for 10 minutes to dissociate cells from surrounding matrix and create a single-cell suspension. Cells were resuspended in media containing 20% serum and washed twice with PBS to remove residual Matrigel.

For coculture assays, shATG7-1 was expressed in HRAS^V12 cells stably expressing pBABEhygro–GFP. This GFP-labeled “target” cell line was then cultured in isolation or combined with unlabeled (pBABEhygro) HRAS^V12 shCNT cells at a ratio of 3:1, with total cell number kept constant at 7,500 cells per well. For conditioned media experiments, HRAS^V12 shATG-expressing cells were grown in 3D culture for 3 days; subsequently, the media was replaced with conditioned media harvested from BABE or HRAS^V12 shCNT MCF10A cells grown in 3D culture for 6 to 8 days. When indicated, 25 μg/mL anti–IL-6 function–blocking or IgG isotype control antibody was added to the conditioned media.

Cells were grown to confluence in 3.5-cm dishes and incubated overnight in assay media lacking EGF for MCF10A cells or DMEM + 2% FBS for MDA-MB-231 cells. Wound healing was performed in the presence of 2 μg/mL mitomycin C (Sigma). Cells were wounded with a 200-μL pipette tip and imaged at the time of wounding (0 hours) and the indicated time points. Average wound widths were measured at each time point, and decreases in wound width were calculated by subtracting the average width at the final time point from the average width at 0 hours using MetaMorph Software (v6.0).

Cells were starved overnight in assay media lacking EGF and then plated at 1.0 × 10⁵ in the top chamber of an 8-μm Transwell filter in assay media lacking EGF. The bottom chamber was filled with assay medium containing 5 ng/mL EGF. Cells were allowed to migrate for 24 hours, after which the top of each filter was cleared of cells. Cells attached to the bottom of the filter were fixed and stained with crystal violet. Crystal violet was extracted with 10% acetic acid and the absorbance was measured at 600 nm.

For experimental metastasis assays, cells were infected with pHIV-ZsGreen (Addgene; plasmid 18121). A total of 1.0 × 10⁶ HRAS^V12 shCNT, shATG7-1, and shATG12 cells stably expressing ZsGreen were injected into the tail vein of nonobese diabetic/severe combined immunodeficient (NOD/SCID) mice. After 140 days, whole lungs were fixed and imaged to detect the number of ZsGreen-positive foci per lung. All animal experiments were conducted in accordance with approved protocols of the University of California, San Francisco (UCSF; San Francisco, CA) Institutional Animal Care and Use Committee (IACUC).

Day 5 3D cultures were washed twice with PBS and cultured for 18 hours in serum-free media. Conditioned medium was collected, and total protein levels were determined by BCA assay (Thermo Scientific) to normalize samples. IL-6 levels were measured using the Quantikine High Sensitivity ELISA Kit (R&D Systems).

Each experiment was repeated at least three independent times. GraphPad Prism software (v5.0b) was used for generation of graphs and statistical analyses. P values were determined by a Student t test or ANOVA as stated.

J. Debnath has received honoraria from the Speakers' Bureaus of Amgen and Novartis. No potential conflicts of interest were disclosed by the other authors.

Conception and design: R. Lock, C.M. Kenific, J. Debnath

Development of methodology: R. Lock, C.M. Kenific, J. Debnath

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R. Lock, C.M. Kenific, A.M. Leidal, E. Salas, J. Debnath

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R. Lock, C.M. Kenific, A.M. Leidal, E. Salas, J. Debnath

Writing, review, and/or revision of the manuscript: R. Lock, A.M. Leidal, J. Debnath

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R. Lock

Study supervision: J. Debnath

This study was supported by the NIH (R01CA126792 and CA126792-S1), the Department of Defense Breast Cancer Research Program (W81XWH-11-1-0130 and W81XWH-12-1-0505), the Samuel Waxman Cancer Research Foundation, and the California Tobacco-Related Disease Research Program (18XT-0106) to J. Debnath. R. Lock was supported by a DOD BCRP Predoctoral Award (W81XWH-08-1-0759) and C.M. Kenific by a Genentech Fellowship, a University of California CRCC Fellowship, and NIH F31CA167905.