Programming of macrophages by apoptotic cancer cells inhibits cancer progression through exosomal PTEN and PPARγ ligands

Apoptotic cell clearance by phagocytes is essential in tissue homeostasis. We demonstrated that conditioned medium (CM) from macrophages exposed to apoptotic cancer cells inhibits epithelial-mesenchymal transition (EMT), migration, and invasion of cancer cells with the acquisition of cancer-stem–like traits. Apoptotic 344SQ (ApoSQ) cell-induced PPARγ activity in macrophages caused increased PTEN levels, secreted in exosomes. ApoSQ-exposed CM from PTEN knockdown cells failed to enhance PTEN in recipient 344SQ cells, restore cellular polarity, and exert anti-EMT and anti-invasive effects. The CM which deficient of PPARγ ligands could not reverse the suppression of PPARγ activity and PTEN and consequently failed to the prevent EMT process. Moreover, single injection of ApoSQ cells inhibited lung metastasis in syngeneic mice with enhanced PPARγ/PTEN signaling both in tumor-associated macrophages and tumor cells. PPARγ antagonist GW9662 reversed PTEN signaling and anti-metastatic effect. Thus, apoptotic cancer cell therapy may offer a new strategy for the prevention of metastasis.


Introduction
Metastasis is a complex multistep process of cancer cell dissemination that is extremely difficult to treat. The outcome of cancer patients with metastatic disease has not improved in the past 30 years, in spite of the development of targeted therapies (Tevaarwerk et al., 2013). In one working hypothesis, metastasis is initiated by tumor cells that undergo epithelial-to-mesenchymal transition (EMT) in response to extracellular signals, leading to loss of polarity, detachment from neighboring cells, increased motility, invasion into surrounding matrix, and resistance to standard cytotoxic chemotherapy (Singh and Settleman, 2010).
PTEN (phosphatase and tensin homolog on chromosome ten), a powerful and multifaceted suppressor, is mutated in multiple types of cancer (Li et al., 1997) and has both phosphatase-dependent and -independent roles. PTEN encodes a dual-specificity phosphatase whose primary substrate is phosphatidylinositol 3,4,5 triphosphate (PIP3) (Cully et al., 2006). PTEN antagonizes phosphoinositide 3-kinase (PI3K) signaling and thereby affects several cellular processes, including growth, proliferation, and survival (Cantley, 2002, Wang et al., 2015. A number of clinical studies demonstrate that PTEN suppression or loss in advanced stage disease contributes to tumor invasion and metastasis (Wikman et al., 2012, Mulholland et al., 2012. PTEN knockdown in human colon cancer cells or prostate cancer cells leads to EMT induction, associated with invasion and metastasis (Bowen et al., 2009). In mice, PTEN loss results in neoplastic growth, in tumors and in the tumor microenvironment (Podsypanina et al., 1999, Trimboli et al., 2009. Two recent studies have found, surprisingly, that PTEN can be secreted unconventionally via exosome formation, and that the translational variant PTEN-long form can be secreted via an unknown mechanism from donor cells and enter neighboring cells (Hopkins et al., 2013, Putz et al., 2012. Like cytoplasmic PTEN, secreted PTEN has lipid phosphatase activity and 4 antagonizes PI3K signaling in target cells, inducing tumor regression. Peroxisome proliferator-activated receptor gamma (PPAR) is a potential PTEN transcription factor (Patel et al., 2001); its activation through ligands increases functional PTEN protein expression in various cancer cell lines, subsequently inhibiting Akt phosphorylation and cellular growth (Teresi et al., 2006, Zhang et al., 2006. Several in vivo studies have demonstrated that genetic alterations of PPAR can promote tumor progression (Yu et al., 2010, Shen et al., 2012. These studies suggest the importance of PPAR/PTEN signaling in cancer prevention. Apoptotic cell clearance by tissue macrophages and non-professional phagocytes (efferocytosis) is essential in tissue homeostasis, immunity, and inflammation resolution.
High levels of cell death can occur within the tumor environment, and clearance mechanisms for dying tumor cells can profoundly influence tumor-specific immunity. Previously, we demonstrated that in vitro and in vivo exposure of macrophages to apoptotic cells inhibits TGF-1 or bleomycin-induced EMT in lung alveolar epithelial cells (Yoon et al., 2016).
However, the effects of efferocytosis in the multistep process of cancer cell dissemination, leading to cancer metastasis, have not been studied thus far. Here, using in vitro 2D and 3D culture systems, we investigated whether the interaction of macrophages with dying cancer cells inhibits EMT in various epithelial cancer cells, and decreases cancer cell migration and invasiveness. We demonstrated that PTEN secretion in exosomes and the PPAR ligands from macrophages exposed to apoptotic cancer cells block the multistep metastatic process.
Furthermore, we provided in vivo evidence that subcutaneous injection of apoptotic cancer cells inhibits the numbers of visible lung metastases of the primary subcutaneous tumor via PPAR/PTEN signaling.

Interaction between macrophages and apoptotic cancer cells inhibits TGF-1-induced EMT in cancer cells
To determine whether programming macrophages by apoptotic cancer cells inhibits EMT in epithelial cancer cells, 344SQ cells were treated with conditioned medium (CM) from RAW cells exposed to either apoptotic 344SQ murine lung adenocarcinoma cells (ApoSQexposed CM) or necrotic cells (NecSQ-exposed CM), along with TGF-1. ApoSQ-exposed CM inhibited TGF-1-induced EMT, based on morphological cellular alterations ( Figure 1A), and EMT marker mRNA (Supplementary figure 1A) and protein ( Figure 1B) expression profiles. In contrast, NecSQ-exposed CM did not exhibit anti-EMT effects. We confirmed the anti-EMT effects of various types of apoptotic cancer cell-exposed CM in the human non- TGF--induced EMT is achieved through the well-orchestrated actions of the Snai, 6 ZEB, and Basic helix-loop-helix transcription factor families (Xu et al., 2009). We observed that ApoSQ or ApoA-exposed CM markedly inhibited TGF-1-induced Snai1/2, Zeb1/2, and Twist1 mRNA expression ( Figure 1D; Supplementary figure 3A), whereas control, NecSQ-, or NecA-exposed CM did not. Intracellular signaling studies (Supplementary figure 3B) show that Smad-dependent TGF- signaling and the ERK pathway were not affected (Supplementary figure 3C-E). However, ApoSQ-exposed CM partially blocked Smadindependent TGF- signaling, including the p38 MAP kinase and Akt pathways, in 344SQ cells (Supplementary figure 3F, G).

Interaction of macrophages with apoptotic cancer cells inhibits TGF-1-induced cancer cell migration and invasion
Acquisition of mesenchymal state by malignant cancer cells is associated with decreased cell-cell adhesion, and increased migratory and invasive properties, which are crucial for metastasis (Lamouille et al., 2014). These processes are consistent with the acquisition of a cancer stem-like cell phenotype also known as 'stemness' or cancer stem cell characteristics (Brabletz et al., 2005). Our data show that ApoSQ-or ApoA-exposed CM prevented TGF-1-induced cancer cell migration and invasion ( Figure 1E,F; Supplementary figure 3H,I), whereas control, NecSQ-, or NecA-exposed CM did not. In addition, TGF-1induced gene-level enhancement of cancer stem-like cell markers, such as CD90, Oct4, CD44, CD133, or ALDH1A, was reduced by treatment with ApoSQ or ApoA-exposed CM (Supplementary figure 3J,K). Moreover, 3D Matrigel culture confirmed the anti-invasive effect of ApoSQ-exposed CM, which caused 344SQ cells to recover their lost polarity and acinuslike colonies, and caused invasive projection by TGF-1 ( Figure 1G). 7 PPAR-dependent PTEN secretion in exosomes from macrophages exposed to apoptotic cancer cells PPAR expression and activity were enhanced in macrophages exposed to apoptotic cells in vitro and in vivo (Freire-de-Lima et al., 2006, Yoon et al., 2015. We hypothesized that PPAR-dependent PTEN production in macrophages in response to apoptotic cancer cells plays a crucial role in the anti-EMT effect in cancer cells. To prove this, we examined PPAR expression and activity, and PTEN production, in ApoSQ-exposed RAW cells. PPAR mRNA and protein expression, and its activity, were markedly enhanced ( Figure 2D) reversed the PTEN mRNA and protein abundance in ApoSQ-exposed RAW cells. Similar PPAR-dependent PTEN induction was shown in naive BMDMs and M2-phenotype BMDMs exposed to ApoSQ (Supplementary figure 4G,H). These data suggest that this important protein is transcriptionally upregulated by enhanced PPAR expression and activation in macrophages, upon stimulation with apoptotic cancer cells.
Interestingly, GW9662 treatment in RAW cells significantly reversed the anti-EMT effects of ApoSQ-exposed CM (Supplementary figure 4I-L). PPAR-dependent PTEN production in this experimental context might be a candidate for the acquisition of these anti-EMT and anti-invasive effects of the CM, if PTEN can be secreted in exosomes and 8 secreted PTEN internalized by recipient cells, with consequent functional activity. To confirm this assumption, we first examined whether the enhanced PTEN protein levels could be secreted in exosomes from macrophages. Transmission electron microscopy revealed microvesicles in RAW cells treated with ApoSQ (Supplementary figure 4M). In addition to ApoSQ-induced PTEN expression in BMDM whole-cell lysates, we observed strong PTEN expression in ApoSQ-exposed CM ( Figure 2E). Moreover, exosomes were isolated by sequential ultracentrifugation (Putz et al., 2012) and used for western blotting analysis using anti-PTEN and exsosomal marker CD63 antibodies. Importantly, PTEN was recovered in the insoluble fraction with the exosomal marker CD63, confirming the presence of PTEN in ApoSQ-exposed CM, whereas GW9662 treatment reduced PTEN abundance ( Figure 2F).

344SQ cells
To ascertain entry of PTEN-bearing exosomes secreted from macrophages to the recipient cancer cells, CM from human macrophage cells overexpressing GFP-PTEN exposed to ApoA was subjected to sequential ultracentrifugation for the harvesting of exosomes. We confirmed the presence of exosomal PTEN in the insoluble fraction through western blotting analysis using anti-GFP and CD63 antibodies ( Figure 2G). After treating cells, such as 344SQ and A549 cells, with the harvested exosomes for 24 h, GFP fluorescence was detected in the cells using confocal microscopy ( Figure 2H), indicating uptake of exosomal GFP-PTEN. This observation was confirmed using western blotting with anti-GFP antibodies ( Figure 2I). PTEN protein levels in recipient 344SQ cells increased immediately and remained so up to 24 h, and basal Akt phosphorylation decreased reciprocally over 6 h, following treatment with ApoSQ-exposed CM ( Figure 2J). Moreover, PTEN abundance was inversely proportional to the dilution of ApoSQ-exposed CM ( Figure 2K), although its mRNA abundance was not affected until 24 h after ApoSQ-exposed CM treatment in the absence or presence of TGF-1 ( Figure 2L). TGF-1 itself did not affect basal PTEN protein abundance for 24 h, and thereafter caused it to decrease in 344SQ cells (Supplementary figure 4N).
Reciprocal to PTEN abundance, TGF-1-induced Akt phosphorylation decreased until 24 h after ApoSQ-exposed CM treatment ( Figure 2M). However, PTEN enhancement, and the effective reduction of Akt phosphorylation or p38 MAP kinase, were not seen with ApoSQexposed CM from PTEN knockdown RAW cells (Supplementary figure 4O; Figure 2N,O).
These data suggest that entered PTEN is functionally active to alter basal Akt signaling and TGF-1-induced non-Smad signaling.
We next investigated whether PTEN is recruited to the plasma membrane after treatment with ApoSQ-exposed CM to sustain cell polarity. Twelve hours after ApoSQexposed CM treatment, PTEN protein was enhanced in the plasma membrane-enriched fraction in 344SQ cells ( Figure 3A). Pretreatment with the PTEN-selective inhibitor SF1670, which elevates intracellular PIP3 signaling (Li et al., 2014), reversed the inhibition of basal Akt phosphorylation in the plasma membrane-enriched fraction. These data suggest that PTEN is recruited to the plasma membrane and functionally inhibits Akt activation.
Loss of PTEN function prevents normal apical surface and lumen development in 3D-culture (Martin-Belmonte et al., 2007). Thus, to evaluate the role of PTEN in cell polarity, 344SQ cells grown in 3D Matrigel were treated with CM from PTEN knockdown RAW cells exposed to ApoSQ, stained with anti--catenin (green), and examined by confocal microscopy. Treatment with ApoSQ-exposed CM from control siRNA-transfected macrophages prevented TGF-1-induced interference, with the formation of polarized acinar structures by 344SQ cells during relatively early exposure (12 h after TGF-1 treatment) ( Figure 3B). ApoSQ-exposed CM from PTEN knockdown macrophages did not exert this inhibitory effect. Moreover, TGF-1-induced disruption of cell-to-cell contacts was prevented by ApoSQ-exposed CM treatment, whereas CM from PTEN knockdown macrophages suppressed this preventive effect ( Figure 3C). Similarly, pretreatment of 344SQ cells with SF1670 inhibited ApoSQ-exposed CM-induced restoration of cellular polarity in 3D Matrigel culture ( Figure 3D) and cell-to-cell contact, as evidenced by confocal microscopy after anti--catenin staining (red) ( Figure 3E).
Next, we examined whether PTEN contributes to the late-phase anti-EMT and antiinvasive effects of ApoSQ-exposed CM. PTEN knockdown in RAW cells substantially inhibited ApoSQ-exposed CM-induced EMT marker changes ( Figure 3F, G), and invading cell number reduction ( Figure 3H), 48 h after TGF-1 treatment. Similarly, PTEN inhibition by SF1670 in 344SQ cells reversed the anti-EMT and anti-invasion effects of ApoSQ-exposed CM (Supplementary figure 3P; Figure 3I-K). These data indicate that PTEN or its signaling mediates the anti-EMT and anti-invasive effects of ApoSQ-exposed CM in the late phase in TGF-1-stimulated 344SQ cells.

Exogenous treatment of cancer cells with ligands inhibits EMT via enhanced PPAR/PTEN signaling
To confirm that 15-HETE, lipoxin A4, and 15d-PGJ2 act in a paracrine manner to induce anti-EMT effects through enhanced PPAR/PTEN signaling, we investigated the effects of these soluble mediators on 344SQ cells at basal (80, 73, and 73 pg/ml, 12 respectively) and stimulatory (258, 422, and 226 pg/ml, respectively) concentrations.
Stimulatory concentrations of all these ligands combined enhanced PPAR activity after 36 h, whereas basal concentrations exerted no effect (Supplementary figure 6A). As expected, each ligand partially inhibited late-phase TGF-1-induced EMT process at its stimulatory, but

PTEN sources in recipient 344SQ cells include internalized PTEN and liganddependent PPAR activity
We further investigated sources of PTEN signaling in 344SQ cells over time following ApoSQ-exposed CM treatment, using pharmacological strategies. The effect of ApoSQ-exposed CM on PTEN mRNA abundance at 48 h in the absence (Supplementary figure 7A) or presence (Supplementary figure 7B) of TGF- stimulation was inhibited by GW9662 treatment in 344SQ cells, but not in RAW cells, indicating PPAR-dependent PTEN induction in 344SQ cells. However, enhanced PTEN protein abundance in 344SQ cells in the presence of TGF-1 at 12 h after ApoSQ-exposed CM treatment was reduced by GW9662 treatment in RAW cells, but not in recipient 344SQ cells, indicating that the PTEN is originated from PPAR-dependent PTEN induction in RAW cells (Supplementary figure   7C). Associated with PTEN reduction in 344SQ cells, GW9662 treatment in RAW cells, but not in 344SQ cells, could inhibit cell polarity restoration by ApoSQ-exposed CM at 12 h 13 (Supplementary figure 7D). However, ApoSQ-exposed CM from RAW cells treated with PD146176 or transfected with L-PGDS-siRNA did not alter PTEN protein abundance in 344SQ cells at 12 h after TGF-1 stimulation (Supplementary figure 7E,G) and accordingly, had no effect on retaining cell polarity (Supplementary figure 7F, H). These data indicate that PTEN sources in recipient 344SQ cells include internalized PTEN in the early phase, and ligand-dependent PPAR activity in the late phase, after ApoSQ-exposed CM treatment.

Apoptotic cell treatment suppresses metastasis and enhances PPAR/PTEN signaling in vivo
To explore the effects of apoptotic cancer cells in mouse metastasis models, we injected syngeneic (129/S) immunocompetent mice subcutaneously with highly metastatic 344SQ cells and allowed them to grow for 6 weeks ( Figure 5A  Immunohistochemistry of serial sections of primary tumor tissue confirmed enhanced PPAR (green, Figure 7A,B), PTEN (green, Figure 7E,F), and CD36 expression (red, Figure   7H,I) upon ApoSQ injection. In particular, PPAR ( Figure  Of note, in the ApoSQ injection group, we observed PTEN-bearing exosomes in the tumor sections using confocal microscopy ( Figure 8A). In addition, increased PTEN expression was detected in the harvested exosomes from serum and ascitic fluid ( Figure   8B,C).
To confirm PPAR-dependent PTEN expression and concurrently mediating antmetastasis effect, the PPAR antagonist GW9662 (1mg/kg/d) was i.p. administered for 4 weeks, beginning one day before ApoSQ injection. GW9662 treatment reversed reduction of 15 metastatic incidence by ApoSQ injection ( Figure 9A). Interestingly, the enhancement of PTEN and CD36 and the reduction of Snai1 and Zeb1 mRNA expression in tumor tissue by ApoSQ injection were reversed by GW9662 treatment ( Figure 9B). Moreover, ApoSQinduced enhancement of PTEN protein expression and reduction of Akt phosphorylation were also reversed by GW9662 treatment ( Figure 9C). This inhibitor administered with buffer had no effects.

Discussion
EMT activation has previously been proposed as the critical mechanism in malignant phenotype acquisition by epithelial cancer cells (Thiery, 2002); we now propose that the interaction between macrophages and apoptotic cancer cells could provide an anti-cancer microenvironment to inhibit EMT and the multistep process of cancer cell dissemination. Our in vitro data demonstrate that ApoSQ-exposed CM from RAW cells and primary mouse BMDMs inhibited TGF-1-induced EMT in 344SQ cells. In addition to murine lung adenocarcinoma cells, the interaction of macrophages with various types of apoptotic human cancer cells, such as non-small cell lung, breast, colon, and prostate cancer cells, but not necrotic cells, results in the inhibition of EMT marker changes. These data clearly suggest that this anti-EMT effect is universal and specific. Similar to wild-type mouse and human macrophages, CM from M2-like BMDMs sharing properties TAM phenotype (Sica et al., 2006), and blood MDMs of lung adenocarcinoma patients exposed to apoptotic cancer cells, show anti-EMT effects. These data suggest that macrophages under normal or cancer circumstances may consistently have the ability to prevent EMT in response to apoptotic cancer cells.
Our data suggest that downregulation of Smad-independent signaling, including p38 MAP kinase and Akt pathways, by ApoSQ-exposed CM inactivates transcription factors that bind to the Snai1/2, Zeb1/2, and Twist1 promoters in the tumor microenvironment. We found that apoptotic cancer cell-exposed CM inhibits migration and invasion, as well as cancer stem-like phenotype acquisition, in cancer cells. Moreover, anti-invasive effect of the ApoSQexposed CM was confirmed using 3D Matrigel culture. These findings provide the new insight that macrophages exposed to apoptotic cancer cells might create a tumor microenvironment preventing metastatic processes. This is the first report of PPAR activity-dependent PTEN induction in macrophages exposed to apoptotic, but not necrotic, cancer cells. Recently, two different groups demonstrated that PTEN secreted via exosome formation, or PTEN-Long via an unknown mechanism, can be internalized by recipient cells (Hopkins et al., 2013, Putz et al., 2012. Surprisingly, we found enhanced canonical PTEN expression in PTEN immunoprecipitates from ApoSQ-exposed CM, and confirmed the secretion of this protein via exosome formation from multivesicular bodies. PTEN secretion in exosomes by human macrophages overexpressing GFP-PTEN in response to ApoA was also verified. Moreover, PTEN-bearing exosomes appear to be uptaken by recipient cancer cells. The way individual vesicles interact with recipient cells is still not known, and has been proposed to involve binding at the cell surface via specific receptors, internalization by a variety of endocytic pathways or micropinocytosis, and/or fusion with plasma membrane or with the limiting membrane of internal compartments (Thery, 2011). Accordingly, PTEN abundance in recipient 344SQ cells was enhanced immediately, and remained until 24 h after treatment with ApoSQexposed CM, without changes in PTEN mRNA expression; Akt phosphorylation decreased reciprocally. Interestingly, ApoSQ-exposed CM from PTEN knockdown RAW cells failed to enhance PTEN abundance and reduce TGF-induced p38 MAP kinase and Akt phosphorylation. Taken together, these data indicate that enhanced PTEN protein levels in 344SQ cells do not originate from early-phase transcriptional induction, but from its internalization into recipient cells with intact lipid and possibly protein phosphatase activity (Tibarewal et al., 2012).
PTEN functions in a spatially restricted manner, which may explain its involvement in forming PIP3 gradients, necessary for generating and/or sustaining cell polarity in epithelial tissues (Martin-Belmonte et al., 2007). Accumulating evidence indicates that loss of cellular polarity and tissue architecture can drive tumor progression (Wodarz and Nathke, 2007). We observed enhanced PTEN recruitment to the cancer cell plasma membrane 12 h 18 after the addition of ApoSQ-exposed CM, and concomitantly reduced Akt phosphorylation in the plasma membrane-enriched fraction. Moreover, our data from the experiments using PTEN knockdown in RAW cells or PTEN signaling inhibition in 344SQ cells highlight that signaling through internalized PTEN mediates prolonged anti-EMT and anti-invasion effects, controlling early cell polarity and integrity by preventing the dissolution of cell-cell contacts.
On the other hand, based on the prolonged effects of restored PTEN mRNA expression over 48 h following ApoSQ-exposed CM treatment, we propose that PTEN signaling in recipient 344SQ cells may originate from a different source, although the PTEN internalization rate, half-life, and stability of internalized PTEN were not estimated (Vazquez et al., 2000). Notably, with regard to PPAR activation over 72 h after treatment with ApoSQexposed CM, 344SQ cells demonstrated a striking resemblance to cells with PTEN mRNA expression, with the dependence on PPAR activity. These data support the novel insight that enhanced late-phase PTEN mRNA expression may be induced mainly through PPARdependent transcriptional upregulation in recipient cancer cells.
We observed enhanced secretion of PPAR ligands in ApoSQ-exposed CM, but not in viable or necrotic 344SQ-exposed CM. Interestingly, ApoSQ-exposed CM deficient in these PPAR ligands partially failed to reverse the reduction of PPAR mRNA and activity, and PTEN mRNA and protein levels, in recipient 344SQ cells, and consequently, EMT under TGF-1 stimulation was not effectively prevented. These data suggest that ligand-dependent PPAR/PTEN signaling in 344SQ cells also mediates the anti-EMT effects of ApoSQexposed CM. Supporting this hypothesis, the exogenous addition of these lipid mediators into 344SQ cells partially reversed TGF1-induced reduction of PPAR mRNA expression and activation, and PTEN mRNA and protein expression, and concomitantly inhibited EMT.
Unlike ApoSQ-exposed CM, these mediators did not affect early-phase TGF1-induced Akt and p38 MAP kinase phosphorylation, indicating that their anti-EMT effect bears no relation to early signaling events.
We identified the original sources of PTEN signaling in 344SQ cells over time after ApoSQ-exposed CM treatment: internalized PTEN (predominant in the early phase), and ligand-dependent PPAR signaling (predominant in the late phase). Furthermore, cell polarity maintenance in the early phase after treatment with the CM is attributed primarily to internalized PTEN, as inhibiting PPAR activity in RAW cells suppresses retaining polarity of 344SQ cells at 12 h after ApoSQ-exposed CM treatment. However, inhibited PPAR ligand production in macrophages could not suppress this early effect of ApoSQ-exposed CM.
Increasing evidence indicates that PTEN loss triggers EMT in many cancer cell types, and consequently promotes invasion and metastasis in various cancers (Wang et al., 2015, Mulholland et al., 2012, Bowen et al., 2009. Intraperitoneally injected PTEN-Long (a translational variant of PTEN) led to tumor regression in several xenograft models, dependent on PTEN-Long phosphatase activity (Hopkins et al., 2013). Notably, PTEN deletion in stromal fibroblasts accelerated the initiation, progression, and malignant transformation of mammary epithelial tumors (Trimboli et al., 2009). In the present study, a single administration of ApoSQ around the lesion two days after 344SQ cell injection into syngeneic mice diminished number of metastatic nodules and lung metastasis incidence after 6 weeks, but caused no significant changes in primary tumor size, indicating the antimetastatic effect of ApoSQ injection in vivo. However, some mice treated with ApoSQ developed lung metastases after 6 weeks of treatment with 344SQ cells. ApoSQ use should be modulated; more injections, modified injection timing, combined therapy with efferocytosis-stimulating agents, or PTEN-bearing exosome therapy with PPAR ligands might be needed. Importantly, a single ApoSQ injection leads to enhanced induction of PPAR and PTEN mRNA and protein expression, and a reciprocal reduction of 20 phosphorylated Akt, as well as mRNA levels of Snai1 and Zeb1, within primary tumor tissue. In summary, we propose that PTEN secretion in exosomes from macrophages

Antibodies
The antibodies used for the Western blotting and immunofluorescence are listed in Supplementary table 1.

Cell lines, primary cells, and culture
Murine RAW 264.7 cells and human cancer cell lines were obtained from ATCC (American Type Culture Collection). 344SQ cells (gift from Dr. Kurie) (Gibbons et al., 2009)

Conditioned medium
Murine macrophages (RAW, BMDM and M2-like cells) or human blood MDM were plated at 5 × 10 5 cells/ml and grown in suitable medium (refer to Cell cultures) at 37°C and 5% CO2.
After overnight incubation, the cells were then serum-starved with X-VIVO 10 medium (04-380Q, Lonza) for 24 h before cell stimulations. For the stimulation, the culture medium was replaced with X-VIVO 10 containing apoptotic or necrotic cancer cells (1.5 × 10 6 cells/ml).
After 24 h, supernatants were harvested by centrifugation and used as the conditioned medium for stimulation of target cancer epithelial cells (5 × 10 5 cells/ml).

Blood samples from patients
Lung cancer patients and healthy controls were included in the study after informed consent under protocols approved by the Institutional Review Board of Ewha Womans University, School of Medicine. A total of three health control (one male, two females) and three nonsmall cell lung cancer patients without anticancer drugs (one male, two females) were used in the experiments depicted in Figure 1C, 2C and Supplementary figure 1J,K. Human monocytes were collected from 20 ml blood by Ficoll-Histopaque density gradient centrifugation (Repnik et al., 2003). Purified monocytes were grown in RPMI containing 10% human AB serum for 8 days. The confirmation of the monocyte differentiation into macrophages (MDM) was done by confocal microscopy with anti-F4/80.

Induction of cell death
Cancer epithelial cell lines were exposed to ultraviolet irradiation at 254 nm for 10 min 24 followed by incubation in RPMI-1640 with 10% FBS for 2 h at 37C and 5% CO2. Evaluation of nuclear morphology using light microscopy on Wright-Giemsa-stained samples indicated that the irradiated cells were approximately apoptotic (Byun et al., 2014). Lysed (necrotic) cancer cells were obtained by multiple freeze-thaw cycles (Fadok et al., 2001). Apoptosis and necrosis were confirmed by Annexin V-FITC/propidium iodide (BD Biosciences, San Jose, CA) staining followed by flow cytometric analysis on a FACSCalibur system (BD Biosciences) (Byun et al., 2014).
Immunoprecipitation 60 ml conditioned medium was prepared from BMDMs that had been non-or apoptotic 344SQ cell-stimulated at 1 × 10 6 per ml. In the case of unconventional secretion of PTEN (Chua et al., 2014), the medium was diluted 1:1 with 2X exosome lysis buffer (4% SDS, 2% Triton-X100, 0.1 M Tris pH 7.4 and 2X protease inhibitors) (de Jong et al., 2012) and lysed for 1 hour at 4°C. The resulting medium-lysis buffer mixture was filtered through 0.22 micron filter (Macherey-Nagel) and divided into 15 ml vials. 25 μl of anti-PTEN (138G6, Cell Signaling) was added to a mixture vial and reacted for 4 h with rotation at 4°C.
Immunocomplexes were then precipitated with 100 μl (50% slurry) of Protein A/G Sepharose (BioVision Inc). Pull-down beads prebound immunocomplex were added into the new tube and incubated for 4 h with rotation at 4°C repeatedly for the rest of mixture vials, and washed for times with IP wash buffer (25 mM HEPES pH7.4, 1 M Nacl, 1 mM EDTA, 0.5 % Triton X-100). To avoid overlapping with IgG Heavy chains, the Western blotting for PTEN immunoprecipitates was performed in non-reducing conditions.

25
Standard western blottings were performed using whole cell extracts, (in)soluble fractionates from conditioned media, or immunoprecipitates. The information of antibodies was included in Supplementary table 1.

Real-time quantitative PCR (qPCR)
mRNAs were extracted from cells grown to 80% confluence in triplicate on 6-well plates in the experimental conditions and quantified using Real-Time PCR System (Applied Biosystems, Step One Plus). See Supplementary table 2 for primer sequences of target genes.

Migration and invasion assays
Cell migration and invasion were tested using Transwell chambers (Corning Inc) coated with 10 μg/ml fibronectin and 200 μg/ml Matrigel matrix according to the manufacturer`s instruction, respectively. In brief, pre-incubated cancer cells (5 × 10 4 cells/well for the migration assay and 2 × 10 5 cells/well for the invasion assay) in the conditioned medium from macrophages in the absence or presence of 10 ng/ml TGF-β1 were plated in replicate wells in serum-free RPMI in the upper chambers and in RPMI 1640 supplemented with 10% FBS placed in the bottom wells at 37°C for 16 h migration time or 24 h invasion time. After fixation in 4% paraformaldehyde, the nonmigrated or noninvaded cells on the upper surface of the membrane were scraped off with a cotton swap. The cells on the lower surface were stained using 0.1% crystal violet, and washed with distilled water. Three random microscopic fields (10X magnification) were photographed and counted.

26
Standard 3D culture was performed as described previously (Debnath et al., 2003). Briefly, a single-cell suspension containing 5000 cells/well was plated on top layer of the solidified Growth Factor Reduced Matrigel (Corning Inc) in 8 well plate. The cells in RPMI 1640 with 10% FBS and 2% Matrigel were incubated and the medium was changed every two or three days for a week. After incubation, the cells were treated with the indicated conditioned medium containing TGF-β1 (10 ng/ml) and 2% Matrigel, and then grown for 3 days. Phase contrast images were taken using Eclipse TE-300 microscope (Nikon).

Immunofluorescence
344SQ cells grown on glass coverslips until confluent were fixed with 4% paraformaldehyde (PFA) solution for 8 min at room temperature (RT). For staining of 344SQ acini in Matrigel 3-D culture, paraffin-embedded tumor or frozen skin tissues, formalin fixation was performed at RT for 30 minutes and IF-Wash buffer (0.05% NaN3, 0.1% BSA, 0.2% Triton X-100 and 0.05% Tween-20 in PBS) was used. After fixation, samples were washed three times with wash buffers for 5 min each and permeabilized with 0.5% Triton X-100 in PBS at RT for 5 min. 5% bovine serum albumin in PBS and -containing Mouse IgG Blocking Reagent were used for ICC and IHC, respectively. Subsequently, all slides stained with antibodies were mounted with Vectashield Mounting Medium containing DAPI (Vector Laboratories, Inc), and imaged with a confocal microscope (LSM 800, Carl Zeiss). The antibody information for sources or dilution ratios is described in Supplementary table 1.

Membrane fractionation
Plasma membrane fractions were prepared from each 150 mm dish of 344SQ cells that had been treated with the indicated conditions as described in Figure 3A

Transmission electron microscopy
Conventional TEM sample preparation was done with RAW264.7 macrophages stimulated by apoptotic 344SQ cells. Ultra-thin sections were then imaged with a transmission electron microscope (H-7650; Hitachi) run at an accelerating voltage of 80 kV. iTEM software (Olympus) was used for image acquisition.

PPAR activity assay
PPAR activity was determined in nuclear extracts (8 g) from pharmacological inhibitorpretreated or siRNA-transfected RAW264.7 or 344SQ cells using a TransAM TM PPAR Transcription Factor Assay kit (40196, Activ Motif Inc) according to the manufacturer's instructions.

Exosome purification
Exosomes were isolated from cell culture media by differential centrifugation as described previously (Putz et al., 2015). In brief, supernatant from RAW264.7 cells exposed to apoptotic or necrotic 344SQ cells was subjected to serial centrifugation of 200 g, 20,000 g and 100,000 g for clearance of dead cells, cell debris and non-exosomal fraction. After washing the exosome pellet with ice-cold PBS containing protease inhibitors, the ultracentrifugation (Optima L-100K, Beckman Coulter Inc., Brea, CA, USA) was repeated for 70 min at 100,000 g to get rid of contaminating proteins. The exosome pellet was then resuspended in RIPA buffer containing protease inhibitors. For isolation of exosomes from mouse serum, exosomes were precipitated using Total Exosome Isolation kit (4478360, Thermo Fisher Scientific) was used. Western blotting analysis was performed for identification of exosomal PTEN in conditioned medium or mouse body fluids with anti-CD63 or anti-PTEN.

Generation of stable macrophages overexpressing GFP-PTEN
Standard lentiviral transduction was performed as described previously . In brief, HEK293T cells were co-transfected with pLV-EGFP-PTEN, packaging (psPAX2) and envelope (pCMV-VSV-G) vectors using the TransIT Ⓡ -LT1 Transfection Reagent (MIR 2300, Mirus Bio, Madison, WI, USA) according to the manufacture's instruction, and incubated for overnight. For lentiviral transduction, the reagents were replaced with fresh media and the viral supernatants were collected every 24 h after transfection. Human macrophages (hM) differentiated from THP-1 monocytes (Zhang et al., 2013) were exposed to the supernatant with 8 µg/ml polybrene for 4hr every virus collection time and the infected cells were selected using with 1 g/ml of puromycin-containing media.

Detection of exosomal GFP-PTEN in recipient cells
GFP-PTEN overexpressing THP-1 macrophages were seeded in ten 150 mm tissue culture dishes and grown to 70% confluency. After stimulation with apoptotic A549 cells for 24 h, conditioned media were collected and subjected to serial centrifugation (refer to exosome purification). The exosome pellet was resuspended in 200 l serum free RPMI. Recipient cells (A549 or 344SQ, 1 × 10 5 cells for confocal microscopy or 3 × 10 5 cells for Western blotting analysis) were exposed to 50l of purified exosomes for 24 h. GFP-PTEN in recipient cell were visualized by direct fluorescence, and also detected with anti-GFP using whole cell lysates.

30
The Animal Care Committee of the Ewha Medical Research Institute approved the experimental protocol. Mice were cared for and handled in accordance with the National Institute of Health (NIH) Guide for the Care and Use of Laboratory Animals. Lung cancer metastasis studies with the syngeneic tumor experiments were performed as previously described (Yang et al., 2014). In brief, syngeneic (129Sv) mice (n = 23 per group) of at least 8 weeks old were used for the syngeneic tumor experiments. 1 × 10 6 344SQ cells in singlecell suspension were subcutaneously injected into the right posterior flank. 2 days after the first injection, second injection was performed in a volume of 100 μl of PBS with or without 1 × 10 7 apoptotic 344SQ cells in the same lesion. Mice were monitored daily for tumor growth and sacrificed at 6 weeks after injection. Necropsies were performed to investigate the weights of subcutaneous tumor mass, the lung metastatic status (# of nodules or incidence) and the histological evaluation of formalin fixed, paraffin-embedded, immunofluorescencestained primary tumor. For the inhibition experiments, the selective PPAR antagonist GW9662 (1mg/kg/d) was i.p. administered for 4 weeks, beginning one day before ApoSQ injection. Mice were necropsied 4 weeks after 344SQ cell injection. For the investigation of early activation of dermal macrophages in response to apoptotic cancer cells, 20-week-old wild type of B6129SF2/J mice were injected with 344SQ cells in the same way and killed 6 or 24 h after the second injection. Right flank skin from mouse was isolated and frozen in OCT compound for immunofluorescent staining of the lesional skin tissue.

TAM isolation
Purification of tumor-associated macrophages (TAM) was performed as described previously (Laoui et al.). To obtain single cell suspension from lung tumors of metastatic mouse models with or without injection of ApoSQ (n = 8 per group, refer to Mouse experiments), solid fresh cancer tissues were disaggregated with tumor digestion medium containing Collagenase I,

31
IV and DNase I. After filtration with 70 m sterile nylon gauze, red blood cells were lysed with erythrocyte lysis buffer. Density gradient centrifugation was performed to isolate mononuclear cells from sharp interphase, and sedimented cells (mostly cancer cells) were used for Western blot analysis. From mononuclear cells washed by MACS buffer, TAM were isolated using anti-CD11b + -conjugated magnetic beads and MACS columns (Miltenyi Biotec). Cross-checking for the identification of TAM was done by confocal microscopy with anti-F4/80.

Statistics
Comparisons between 2 mean values ± SEM (control versus experimental) were performed using the two-tailed Student's t test. P values that are less than 0.05 are considered statistically significant. All data were analyzed using Graph Prism 5 software. (GraphPad Software Inc).        Where indicated, GW9662 (1 mg/kg/day, i.p.) or its vehicle (Veh; 2% DMSO in saline) was administered into the left flank one day before SQ injection of ApoSQ in the right skin lesion (n=5 per group). Mice were necropsied 4 weeks after 344SQ cell injection. (A) Lung metastasis incidence (%). qPCR analysis (B) and immunoblot analysis of indicated protein expression (C) in primary tumors. *P < 0.05, **P < 0.01 and ***P < 0.001 (Student's t test).
Data are representative images from five mice per group (C left) or from independent experiments with five mice per group (mean ± s.e.m. in B and C).