Autocrine CSF-1R signaling drives mesothelioma chemoresistance via AKT activation

Clinical management of malignant pleural mesothelioma (MPM) is very challenging because of the uncommon resistance of this tumor to chemotherapy. We report here increased expression of macrophage colony-stimulating-factor-1-receptor (M-CSF/CSF-1R) mRNA in mesothelioma versus normal tissue specimens and demonstrate that CSF-1R expression identifies chemoresistant cells of mesothelial nature in both primary cultures and mesothelioma cell lines. By using RNAi or ligand trapping, we demonstrate that the chemoresistance properties of those cells depend on autocrine CSF-1R signaling. At the single-cell level, the isolated CSF-1Rpos cells exhibit a complex repertoire of pluripotency, epithelial–mesenchymal transition and detoxifying factors, which define a clonogenic, chemoresistant, precursor-like cell sub-population. The simple activation of CSF-1R in untransformed mesothelial cells is sufficient to confer clonogenicity and resistance to pemetrexed, hallmarks of mesothelioma. In addition, this induced a gene expression profile highly mimicking that observed in the MPM cells endogenously expressing the receptor and the ligands, suggesting that CSF-1R expression is mainly responsible for the phenotype of the identified cell sub-populations. The survival of CSF1Rpos cells requires active AKT (v-akt murine thymoma viral oncogene homolog 1) signaling, which contributed to increased levels of nuclear, transcriptionally competent β-catenin. Inhibition of AKT reduced the transcriptional activity of β-catenin-dependent reporters and sensitized the cells to senescence-induced clonogenic death after pemetrexed treatment. This work expands what is known on the non-macrophage functions of CSF-1R and its role in solid tumors, and suggests that CSF-1R signaling may have a critical pathogenic role in a prototypical, inflammation-related cancer such as MPM and therefore may represent a promising target for therapeutic intervention.


Subject Category: Cancer
The colony-stimulating-factor receptor-1 (CSF-1R) is a type III receptor tyrosine kinase primarily reported as responsible for the proliferation, differentiation and survival of the monocytemacrophage cell lineage, as well as their recruitment. 1 The two known ligands of CSF-1R are CSF-1 and the structurally unrelated sibling IL-34 (interleukin-34). Both CSF-1 and IL-34 function through non-competitive binding to CSF-1R by triggering auto-and transphosphorylation of tyrosine residues in the cytoplasmic domain of the receptor, 2 which results in the activation of both phosphatidylinositol 3-kinase (PI3K)dependent and RAS-activated protein kinase-dependent pathways. 3 CSF-1R may function in cancer cells as an oncogene 4 by exploiting Src-dependent and -independent signaling modalities. 5,6 In vivo, elevated expression of CSF-1R has been described in tumors of epithelial origin, such as ovarian, endometrial, breast, prostate, renal and ovarian carcinomas, and has also been found to correlate in some cases with adverse prognosis. [7][8][9][10][11] Studies in breast cancer and renal cell carcinoma suggest that autocrine signaling of CSF-1R may be relevant in vivo as well and modulated by oncogenic signaling (i.e. transforming growth factor-b1 (TGFb1) and EGF, respectively). 11,12 Malignant pleural mesothelioma (MPM) is a neoplastic disease of the pleura with a clear pathogenetic link to chronic inflammation, in most cases due to asbestos exposure. 13 A clinically silent progression and an unusually high resistance to therapy make the prognosis of mesothelioma patients almost invariably poor and survival has changed very little in the past 20 years. 14 MPM is a clinically relevant model to study the relationships between chronic inflammation and chemoresistance, as asbestos exposure of pleural mesothelial cells induces sustained nuclear factor-kB (NF-kB) activation 15 following activation of high-mobility group box 1 signaling. 16,17 Here we show, for the first time, that expression of the CSF-1R identifies a specific cell sub-population exhibiting pluripotency and chemoresistance in both primary MPM cultures and cell lines. A specific gene expression profile and autocrine activation of the CSF-1R signaling characterize the functional uniqueness of these cells, which exhibit increased NF-kB, signal transducer and activator of transcription 3 (STAT3) and AKT (v-akt murine thymoma viral oncogene homolog 1) activation. AKT activation following the forced activation of CSF-1R signaling in untransformed mesothelial cells confers accelerated growth, clonogenicity and resistance to pemetrexed, and leads to increased nuclear levels of functional b-catenin. This phenotype is reversed by inhibition of AKT. This study expands what is known on the function of CSF-1R in nonhematopoietic neoplasms and suggests that interference with the signaling of the CSF-1R pos cells may overcome the chemoresistance of MPM.

Results
Mesothelioma CSF-1R pos cells exist in vivo. To investigate whether CSF-1R is expressed in mesothelioma tissues and if its levels are correlated to the tumor state, we performed a microarray expression analysis of 34 matched pair samples of normal peritoneal tissue versus mesotheliomas (Figure 1a). This revealed an increased mRNA expression of CSF-1R in the mesothelioma tissue as opposed to the mesothelial tissue (Po0.001). This finding did not exclude that tumor-associated macrophages or bone marrow-derived cells may contribute the observed higher CSF-1R expression in the tumor tissue. Therefore, we developed mesothelioma primary cultures (n ¼ 7) from pleural effusions and solid specimens of MPM patients. Fluorescence-activated cell sorting (FACS) staining revealed that all the MPM specimens contained CSF-1R pos cells shortly after harvesting (t ¼ 2-18 h). Staining of the same samples after long-term culturing (2.2±1.1 months: average length of the culture) revealed that the percentage of CSF-1R pos cells was almost unchanged or slightly increased in time (Figure 1b). Costaining with anti-calretinin antibodies indicated that a relatively large fraction of the CSF-1R pos cells within the primary cell cultures expressed the mesothelioma-specific antigen calretinin, 18 thus confirming the mesothelial nature of the CSF-1R pos cells ( Figure 1c). Finally, we assessed the levels of CSF-1 and IL-34, the known ligands for CSF-1R, in the conditioned medium of the primary cultures (t ¼ 45-60). Enzyme-linked immunosorbent assay (ELISA) assay revealed that both CSF-1R ligands were represented in the medium of the primary cultures (in 7/7 cultures and 6/7 cultures for CSF-1 and IL-34, respectively) ( Figure 1d). The absence, at day 60 of  , of detectable, surviving hematopoietic cells (lymphocytes, monocytes, eosinophils and neutrophils), as judged by  the absence of CD3-, CD14-, CD16-, CD19-, CD20-, CD56-and CD45-positive cells within the CSF-1R pos cell fraction (Supplementary Figure 1), suggested that mesothelial CSF-1R pos cells could be responsible for the production of the CSF-1R ligands.
MPM cell lines secrete CSF-1 and IL-34 and express functional CSF-1R. To obtain a suitable experimental system to study the CSF-1R function in mesothelioma cells, we analyzed the expression of CSF-1R and its ligands CSF-1 and IL-34 in a panel of mesothelioma cell lines and an untransformed mesothelial cell line immortalized by h-TERT (LP9) (Figure 2). FACS analysis showed that all the mesothelioma cell lines contained a small sub-population of CSF-1R pos cells (range 2-13%). A small percentage of LP9 cells exhibiting the expression of CSF-1R (o1.5%) was also present in the mesothelial cells (LP9) (Figure 2a). Next, ELISA assay revealed that 7/7 mesothelioma cell lines secreted IL-34 and 6/7 MPM cell lines secreted CSF-1, with the levels of IL-34 being generally higher than those of CSF-1 ( Figure 2b). No detectable IL-34 and very little CSF-1 was produced by the untransformed mesothelial LP9 cells ( Figure 2b). Thus, mesothelioma cell lines expressed all the components of the CSF-1R signaling module, implying active signaling in those cells. To verify this, we treated H-2595 cells with vehicle (phosphate-buffered saline (PBS)), CSF-1 (25 ng/ml) or IL-34 (25 ng/ml). This revealed increased CSF-1R autophosphorylation, as assessed by western blotting with phospho-CSF-1R (Tyr723) antibodies ( Figure 2c  CSF-1R signaling mediates resistance to pemetrexed of MPM cell lines and primary samples. The previous observations suggested that active CSF-1R signaling may transduce prosurvival signals ( Figure 3). Therefore, we investigated the involvement of the CSF-1R/CSF-1/IL-34 system in mediating the resistance of the mesothelioma cells to pemetrexed. First, we observed a significant increase of the CSF-1 and IL-34 mRNA in the cells treated with pemetrexed for 48 h, which matched the increased secretion of the two cytokines in the medium (Figure 3a). Next, FACS staining revealed that the CSF-1R pos cells survived pemetrexed treatment, with no decrease or increase of the receptor-positive cells in the surviving cell fraction (Figures  3b and c). This was not observed in the pemetrexed-treated mesothelial LP9 cells (Figure 3c Figure 2E). Although CSF-1R downregulation affected the viability and clonogenicity of both control (vehicle)-and pemetrexed-treated cells, its action on the pemetrexed-treated cells was much stronger, consistent with a chemosensitizing effect. To broaden the relevance of these findings, we assessed whether CSF-1R pos mesothelioma cells in primary cultures would exhibit similar behavior. When challenged with pemetrexed, all the primary cultures (n ¼ 7) exhibited surviving CSF-1R pos cells, whose number did not significantly change upon treatment ( Figure 3g). Viability of purified CSF1R pos cells from two representative primary cultures was affected by treatment with the CSF-1R decoy (ECD) and further reduced by pemetrexed treatment (Figure 3h). These observations suggest that, like in the cell lines, the CSF-1R expression in primary MPM specimens identifies a chemoresistant cell population sensitive to CSF-1R inhibition, emphasizing the potential for CSF-1R signaling to impinge on the chemoresistance of mesothelioma cells.
FACS-sorted CSF-1R pos cells exhibit a distinct gene expression profile. We investigated whether the repertoire of gene expression exhibited by the CSF-1R pos cells could explain their properties ( Figure 4). In detail, we evaluated whether FACS-purified, CSF-1R pos H-2595 cells would differ from their unsorted counterparts with regard to the expression of pluripotency markers and chemoresistanceassociated genes. Quantitative reverse-transcription PCR (qRT-PCR) analysis revealed enrichment of the CSF-1R pos cells for the expression of NANOG, OCT4, SOX2, ENDOGLIN, c-MYC and NOTCH1 mRNAs. We also found very high levels of ATP-binding cassette transporters, subfamily G, member 2 (ABCG2) mRNA, a membrane drug transporter whose increased expression confers chemoresistant features. 20,21 Epithelial-mesenchymal transition (EMT) genes, such as VIMENTIN, FBN, matrix metalloproteinase-9 (MMP-9) and CD44, and chief EMT modulators, such as SNAI1, SNAI2 and TGF-b1, were also enriched in CSF-1R pos cells. We also found that the CSF-1R pos cells were enriched for IL-1b, shown to modulate CSF-1 and IL-34, in a NF-kB-dependent manner 22 Figure 3B). In addition, according to the enrichment of the sorted cells for aldehydedehydrogenase 1A3 (ALDH1A3) mRNA (Figure 4a), we evaluated the presence of ALDH activity in the CSF-1R pos cells. Briefly, we analyzed the accumulation of a fluorescent ALDH substrate into the cells, by using those treated with a specific ALDH inhibitor (DEAB (diethylaminobenzaldehyde)) as a background control (ALDEFLUOR Kit, STEMCELL Technologies Inc., Vancouver, BC, Canada). This showed that the CSF-1R pos cells were endowed with ALDH activity (Figure 4e, upper panel). ALDH is a detoxifying enzyme whose function has been linked to cancer chemoresistance. 23  CSF-1R pos cells rely on autocrine signaling to AKT for their survival. Both CSF-1 and IL-34 mRNAs were expressed by the ( Figure 5) purified CSF-1R pos cells and secreted in the medium, thereby demonstrating the existence of autocrine production of the ligands (Figure 5a). This was functionally relevant, as western blotting revealed high enrichment of the purified CSF-1R pos cells for phospho-STAT3 (Tyr705), phospho-NF-kB (p-p65-Ser536) and phospho-AKT (Ser473). No differences of the phosphoextracellular signal-regulated kinase 1/2 (ERK1/2) (Tyr704) signal were observed between unsorted and CSF-1R pos cells ( Figure 5b). Next, we investigated the relevance of the autocrine CSF-1 and IL-34 production for the survival of the CSF-1R pos cells. We treated the purified cells with scrambled-, CSF-1-or IL-34-targeting small interfering RNAs (siRNAs) (20 pM, 48 h), which strongly reduced the levels of CSF-1 and IL-34 mRNAs (Supplementary Figures 4A and B). Cell proliferation assays showed that the depletion of either of the ligands by siRNA affected the growth of the CSF-1R pos cells, with the strongest effect following the simultaneous downregulation of both ligands (Figure 5c). Next, we investigated which of the signaling events, downstream to CSF-1R activation, could promote the survival of the CSF-1R pos cells, thereby potentially affecting their  Figure 4C). Staining of the treated cells for SA-b-galactosidase (b-gal) activity showed that the lack of BrdU incorporation was mainly due to senescence, with the LY294002 treatment eliciting the highest number of senescent cells (Figure 4e). This correlated with the upregulation of p21 in the LY294002 CSF-1R pos -treated cells, which was further increased in the pemetrexed-treated samples (Figure 5f). Thus, CSF-1R activation may trigger prosurvival stimuli by reducing spontaneous senescence mainly via AKT activation.  Figure 5B). In addition, the CA-CSF-1Rexpressing LP9 cells were clonogenic and resistant to pemetrexed treatment (Figure 6b). Based on the previous findings ( Figure 5), we evaluated the AKT activation status in CA-CSF-1R-and EV-LP9. Western blotting revealed that, in addition to increased levels of serine-and threoninephosphorylated AKT, the CA-CSF-1R-LP9 cells exhibited increased levels of phosphorylated PDK1(Ser241) and glycogen synthetase kinase-3b (GSK3b) (Ser9), and main upstream and downstream effectors of AKT signaling, respectively (Figure 6c). We also observed increased phosphorylated c-RAF (Ser259), which connects the mitogenactivated protein kinases (ERK1/2) activation to the AKT pathway 31 (Figure 6c). Phosphorylation of GSK3b at serine 9 inactivates this kinase activity and causes stabilization and nuclear localization of b-catenin from the plasma membrane to the cytoplasm and the nucleus, where the latter acts as a transcription cofactor for developmentally regulated genes. 32 According to this, western blotting revealed increased levels of b-catenin in the CA-CSF-1R-LP9 cells (Figure 6c). Additionally, indirect IF indicated increased nuclear localization of b-catenin in the CA-CSF-1R-LP9 cells as compared with their controls (Figure 6d, upper panel). To assess

Discussion
Previous work has reported expression of the CSF-1R and its ligand CSF-1 in solid tissue tumors and cell lines, and in some cases this was shown to be prognostic or to correlate with the in vitro aggressiveness of the cell lines. However, no studies to date have isolated and characterized the CSF-1R-expressing cells and thereby exploring their biological properties at the single-cell level. Thus, we report a novel finding that CSF-1R expression identifies precursor-like, self-renewing mesothelioma cell sub-populations endowed with prominent protumorigenic properties, such as clonogenicity and chemoresistance. In addition, we showed that IL-34, structurally unrelated ligand for the CSF-1R, has a role in the The findings of this work may be clinically relevant. Indeed, we report increased expression of CSF-1R in mesothelioma patient samples as compared with matched normal peritoneal tissues (n ¼ 34). Besides, we showed that CSF-1R pos cells exist and persist within long-term primary cultures (n ¼ 7) and are resistant to pemetrexed in a CSF-1R-dependent manner. The fact that the ex vivo cultures used in this study are virtually devoid of any hematopoietic component and that their strongly similar behavior to long-term established MPM cell lines (n ¼ 7) support what was observed here is mainly a property of stable, CSF-1R-expressing cells of mesothelial nature. Therefore, the experimental model adopted here potentially expands what is known on the tumor-supporting, macrophage-independent functions of the CSF-1R in an autocrine setting. The purified CSF-1R pos cells exhibit enrichment for SOX2, OCT4 and c-MYC, and express low miRNA-let-7d levels. In addition, those cells express NOTCH1 and ENDOGLIN. Interestingly, Notch1 upregulation by the SV40 virus was causally linked to mesothelial cell immortalization. 33 Endoglin promoted resistance of ovarian cancer cells to DNA-damaging agents in vitro and in vivo. 34 In line with this, the expression of ALDH1A3 mRNA and activity and the surface enrichment for drugeffluxing activities, such ABCG2, support the chemoresistant properties of the isolated cell sub-populations. Therefore, expression of all the above factors is likely to confer a precursor-like, chemoresistant status to the CSF-1R pos cells.
In terms of signaling, we provide evidence that activation of AKT is central to the properties of the CSF-1R pos cells, including survival and resistance to pemetrexed. In detail, AKT inhibition indeed counteracted the bypass of senescence induced by the expression of active CSF-1R in pemetrexedtreated mesothelial cells, thereby restoring their chemosensitivity. Interestingly, activation of b-catenin by CSF-1 was shown to promote survival and expression of c-MYC and CyclinD1 in bone-marrow-derived macrophages. 35 Additionally, the increased functional competence of b-catenin upon We have recently shown that functional interaction of STAT3 and NF-kB impinges on mesothelioma resistance to pemetrexed by modulating EMT-associated genes and that butein abated the resistance of mesothelioma cells by decreasing the pSTAT3-NF-kB interaction. 26 This is interesting, given the functional and physical crosstalk between the AKT, STAT3 and NF-kB pathways shown in other experimental systems (such as MYC-driven lymphomas). 36 Thus, an intriguing degree of pathway interaction seems to sustain the expression of pluripotency and EMT factors, ultimately impinging on a chemoresistant phenotype.
An interesting aspect of this study is that interference with the survival of the CSF-1R pos cells (typically representing o10% of the total live cells in unfractionated cultures) abated the resistance of both cell lines and primary samples to pemetrexed at a much greater extent than it would be expected by targeting only the CSF-1R pos cells. It is possible that molecules secreted by the CSF-1R pos cells may confer oncogenic signals to adjacent, non-CSF-1R pos cells. With regard to this, the CSF-1R pos cells have the potential to secrete TGFb1 and IL-1b, potent inducers of EMT, and EMT is a strong determinant of chemoresistance and a prognostic factor for mesothelioma. 37,38 Therefore, the CSF-1R pos cells may act in this regard as tumor-supporting mesenchymal elements. The increase of IL-1b triggered by CSF-1R activation may also be relevant for the induction of CSF-1 and IL-34 expression, as shown in osteoblast studies. 22 Thus, the CSF-1R pos cells can be attractive targets in vivo as interference with their survival broadly affects the protumorigenic features of both mesothelioma primary cultures and cell lines. In this respect, our observations indicate that both CSF-1R and AKT inhibition may be clinically relevant tools to overcome mesothelioma chemoresistance. This may allow expansion of the poor therapeutic portfolio for such a fatal disease. It is also possible that the findings herein may be applied to other inflammation-related neoplasms, of which mesothelioma is a prototype. Cell viability assay. Manual counting and trypan blue dye-exclusion assay were used to assess cell growth and the cell viability. When indicated, cell viability was also assessed by FACS analysis of SYTOX Dead Cell Stain-labeled cells (Life Technologies) according to the manufacturer's instructions.
Clonogenic assays. MPM cell lines were grown to 70% confluence and pulse treated with the indicated drugs or transfected as indicated. After 16 h, cells were detached and seeded at 500-1500 cells per well into six-well dishes (Corning-Costar, Tewksbury, MA, USA) in drug-free media. Fresh media (25%) were added every 3 days. Colonies were stained with crystal violet (Sigma) and colonies (450 cells) were counted after 7-14 days.
MMP-9 quantification. The levels of MMP-9 were quantified in the media conditioned by CSF-1R pos cells and their unsorted counterparts by ELISA, according to the manufacturer's instructions (MMP-9 Human ELISA Kit; Life Technologies).
cDNA synthesis and gene expression. The first-strand cDNA was synthesized according to manufacturer's instructions (High Capacity RNA-to-cDNA Kit; Applied Biosystems, Foster City, CA, USA). Gene expression was measured by real-time PCR using the SYBR Green dye (Applied Biosytems) on a StepOne Instrument (Applied Biosytems). qPCR primers are reported in Supplementary  Table 1. PPIA and 18S were used as an endogenous control. miRNA expression analysis. Let-7d was detected using a specific TaqMan miRNA assay (Applied Biosystems) according to the manufacturer's instructions on a StepOne Instrument (Applied Biosytems). RNU6B was used as an endogenous control.
Lysate preparation and immunoblotting. Cells were lysed in buffer with 50 mM Tris-HCl (pH 8), with 1% NP-40 (Igepal AC-630), 150 mM NaCl and 5 mM EDTA. Extracts were centrifuged at 14 000 Â g for 15 min to remove cell debris. Protein concentrations were determined by colorimetric assay (Bio-Rad, Hercules, CA, USA). Western blotting was performed using the primary antibodies as indicated in Supplementary Table 2. Actin was used as a loading control.
Expression of CSF-1R in malignant pleural mesothelioma M Cioce et al Luciferase reporter assay. To evaluate b-catenin/TCF-4 transcriptional activity, we used a pair of luciferase reporter constructs, TOP-FLASH and FOP-FLASH (Upstate Biotechnology, Lake Placid, NY, USA). Plasmids of TOP-FLASH (with three repeats of the TCF-binding site) or FOP-FLASH (with three repeats of a mutated TCF-binding site) were transfected into cells treated with LY294002 according to the manufacturer's instructions. Luciferase activity was measured using the Dual-Luciferase Reporter Assay System (Promega, Madison, WI, USA), with Renilla luciferase activity as an internal control, 48 h after transfection.
IF microscopy. Briefly, the cells were fixed and permeabilized in paraformaldehyde/methanol, aspecific binding blocked in PBS containing 1% BSA and labeled with the indicated antibodies. The secondary antibody was an Dylight Goat Anti-Rabbit IgG (Abcam, Cambridge, UK). The nuclei were visualized by 4 0 ,6-diamidino-2-phenylindole (DAPI) staining. Merged images were shown.
SA-b-gal staining. SA-b-gal staining was performed on adherent cells according to the manufacturer's instructions (Cell Signaling Technology).
BrdU incorporation assay. Detection of BrdU incorporation was performed by ELISA (BrdU Cell Proliferation Assay Kit; Cell Signaling Technology) according to the manufacturer's instructions.
Flow cytometry. Cells were detached by PBS 1 Â /EDTA 2 mM, fixed with 4% PFA permeabilized (when needed) in 80% methanol (10 min on ice), washed two times with PBS 1 Â and resuspended for antibody staining at 1 Â 10 6 cells per 100 ml in PBS 1 Â /BSA 1%. The list of the primary antibodies, isotypematched controls and secondary antibodies is available in Supplementary Table 2.
Gates were drawn to exclude 499% of aspecific staining (based on the isotypestained samples). Data were acquired using a FACSCalibur Instrument (BD Biosciences, Franklin Lakes, NJ, USA) and analysis was performed by using the Flowing Software 2.0 (Cell Imaging Core, University of Turku, Turku, Finland).
Apoptosis detection. Dead and apoptotic cells were measured by FACS analysis after propidium iodide/Annexin-V staining (Vybrant Apoptosis Assay Kit No. 3; Life Technologies) according to the manufacturer's instructions.
ALDH activity assay. ALDH activity was detected by FACSCalibur instrument (BD Biosciences). ALDEFLUOR Kit (STEMCELL Technologies Inc.) was used according to the manufacturer's instructions. ALDH-positive cells were defined as the cells that displayed greater fluorescence compared with a control staining reaction containing the ALDH inhibitor, DEAB.
Cell sorting. Cells were filtered through a 40 mM mesh to obtain a single-cell suspension and were incubated with primary and secondary antibodies for 45 min on ice. Cell sorting was performed with a MoFlo cell sorter (Dako Cytomation, Fort Collins, CO, USA) or a BD ARIA II (BD Biosciences).
Statistical analysis. The means of each data set were analyzed using a Student's t-test with a two-tailed distribution and assuming equal sample variance.