Tumor–host interactions play a major role in malignancies' initiation and progression. We have reported in the past that tumor cells attenuate genotoxic stress–induced p53 activation in neighboring stromal cells. Herein, we aim to further elucidate cancer cells' impact on signaling within lung cancer stroma. Primary cancer-associated fibroblasts were grown from resected human lung tumors. Lung cancer lines as well as fresh cultures of resected human lung cancers were used to produce conditioned medium (CM) or cocultured with stromal cells. Invasiveness of cancer cells was evaluated by transwell assays, and in vivo tumor growth was tested in Athymic nude mice. We found CM of a large variety of cancer cell lines as well as ex vivo–cultured lung cancers to rapidly induce protein levels of stromal-MDM2. CM of nontransformed cells had no such effect. Mdm2 induction occurred through enhanced translation, was mTORC1-dependent, and correlated with activation of AKT and p70 S6 Kinase. AKT or MDM2 knockdown in fibroblasts reduced the invasion of neighboring cancer cells, independently of stromal-p53. MDM2 overexpression in fibroblasts enhanced cancer cells' invasion and growth of inoculated tumors in mice. Our results indicate that stromal-MDM2 participates in a p53-independent cancer–host feedback mechanism. Soluble cancer-originated signals induce enhanced translation of stromal-MDM2 through AKT/mTORC1 signaling, which in turn enhances the neighboring cancer cells' invasion ability. The role of these tumor–host interactions needs to be further explored.
We uncovered a novel tumor–stroma signaling loop, which is a potentially new therapeutic target in lung cancer and possibly in additional types of cancer.
This article is featured in Highlights of This Issue, p. 799
The role of tumor–host interactions in tumor progression and spread was suggested over a century ago (1). Tumor–host-immune cells' cross-talk has gained most attention in recent years, with the dramatic results of immunotherapy for many types of cancer, although not in all patients (2). Immune cells and tumor vasculature have become real therapeutic targets, enabling some of the most significant advances in cancer therapy. In contrast, cancer-associated fibroblasts (CAF), probably the most abundant and most resilient tumor–stroma cells, are still poorly understood (3). CAFs are a major component of the tumor microenvironment (TME), activated fibroblasts evolving alongside cancer cells and exhibiting several distinct features as compared with normal fibroblasts. These include higher proliferation rate, secretion of growth factors, and extracellular matrix. Importantly, CAFs exhibit tumor-supporting and prometastatic activity, mediated by paracrine influences (4–8), recruitment or modifications of procancer immune-suppressive cells (9–11), extracellular matrix remodeling (12, 13), alterations of the metabolic microenvironment (7), and other less characterized mechanisms (14). CAFs also have a role in resistance to chemotherapy (15), targeted agents (16), and radiotherapy (17). Importantly, CAFs' protumor impact is not universal, as CAF depletion or inhibition in a pancreatic cancer mice model led to an immune-suppressive microenvironment and more aggressive tumors (18, 19). Understanding the molecular and cellular mechanisms involved in CAF-tumor signaling is essential to allow using these signals as therapeutic targets.
Mouse double-minute-2 (MDM2) is mostly recognized as the major regulator of the TP53 tumor-suppressor gene. TP53 incorporates complex signals of DNA damage and other potentially oncogenic types of stress, its activation leading to elimination of transformation-prone cells (20). p53 protein regulation occurs mostly posttranslationally through MDM2-dependent ubiquitination and degradation of p53 (21), MDM2 and its homolog MDMX binding and blocking p53′s transactivation (22), as well as MDM2-mediated p53-neddylation (23). MDM2 gene is a transactivation target of p53 (24), thus forming a p53-MDM2 regulatory node. MDM2 has a critical role as a negative regulator of p53, as demonstrated by embryonic lethality of MDM2-knockout mice that is completely rescued by TP53 knockout (25, 26). However, MDM2 engages in p53-independent oncogenic activity in cancer. MDM2 amplification can be found in p53-mutated tumors, and in other cases involves splice forms that lack the p53-binding site (27, 28). MDM2 overexpression in mice was found to induce sarcomas on the background of p53-null mice (29). MDM2 can promote epithelial-to-mesenchymal transition (EMT), angiogenesis, and metastatic spread in a model of breast cancer harboring mutant p53 (30) and promotes stemness in p53-null pancreatic cancer cells (31). Furthermore, MDM2 enhances translation of XIAP, by binding its mRNA at the internal ribosomal entry site, leading to resistance to radiation (32). Chromatin-repressive regulation by MDM2 in a polycomb repressor complex 2–dependent manner positions MDM2 as regulating stemness both in pluripotent nontransformed cells and in cancer cells, independently of p53 (33). Mechanistically, these effects involve Histone-H2B ubiquitination by MDM2 and repression of transcription (34). Another p53-independent role was recently reported for MDM2 in regulating serine and glycine metabolism and redox homeostasis, being recruited to chromatin at transcription sites of ATF3/4 and inducing genes involved in those pathways, thus improving cancer cell survival under these types of stress (35). MDM2 enhances translation of MYCN (36), upregulation of expression of the p100/NFkB2 gene (37), and induces transcription of the p65/RelA subunit of NF-kB through binding of MDM2 to Sp-1–binding sites (38), all in a p53-independent manner. MDM2 can interfere with the Rb/E2F cell-cycle–inhibitory pathway (39) and increases cell motility by binding and ubiquitinating E-cadherin in breast cancer cells (40). To the best of our knowledge, a role of stromal-MDM2 in tumor progression has not been reported thus far.
In recent years, several studies have reported a paracrine–tumor-suppressor role for p53 within stromal cells, repressing neighboring tumor cells via various mechanisms (41–44). We have speculated that tumor cells would evolve the ability to inhibit p53 in the stroma to overcome this influence (45, 46). Indeed, we have discovered that conditioned medium (CM) of cancer cells (cancer-CM) inhibits the activation of the p53 pathway in cultured fibroblasts by an unidentified soluble factor (45). While investigating p53′s repression mechanism, we found MDM2 translation to be rapidly induced in CAFs treated with cancer-CM, secondary to AKT Serine/Threonine Kinase 1 (AKT1) and mTORC1 activation, in a p53-independent manner. We now report these findings and further focus on stromal-mTOR-MDM2 impact on neighboring cancer cells.
Materials and Methods
Primary cell culture
Tumor and normal lung tissue fragments were sampled from the specimens of lung cancer operations. Dissections were performed at the surgery suite as soon as possible after resection. CAFs and normal-associated fibroblasts (NF) were isolated from the stromal compartments of either lung tumor specimen or normal adjacent lung, respectively. The tumor or the normal tissue was cut under sterile conditions using surgical razor blade to pieces measuring 3 to 5 mm3. Tissue fragments were incubated in a sterile dish with a minimal volume of medium to allow the attachment of the tissue to the bottom of the dish. MEM medium (Biological Industries) was supplemented with 10% FBS (GIBCO, Thermo Fisher Scientific), 2 mmol/L l-glutamine, 1% Eagle nonessential amino acids, 60 μmol/L 2-Marcaptoethanol, and Penicillin–Streptomycin solution (all from Biological Industries). Ten to 14 days later, growing cells were passaged, identified as fibroblasts based on morphology, and confirmed by positive α-smooth muscle actin and fibronectin1 staining (Supplementary Fig. S1). HK3T-immortalized lung CAFs, previously infected with the human telomerase catalytic subunit (hTERT; as described in ref. 47)—alone or with pRetroSuper-p53 shRNA-Blast (shp53 insert sequence: 5′-GACTCCAGTGGTAATCTAC-3′) or pRetroSuper-shmNOXA-Blast (the mouse NOXA gene sequence as shRNA control; shmNOXA insert sequence: 5′-AAGGGACATCTGTACTTCTGG-3′)—were kindly provided by professor Varda Rotter. Mouse embryonic fibroblasts, a generous gift of Professor Moshe Oren, were grown in DMEM medium (Biological Industries) supplemented with 10% FBS and Penicillin–Streptomycin solution.
Cancer cell lines
Cancer cell lines (H1299, MCF-7, H1975, Hepa1-6, PC-9, WM 266-4, A-375, HCC827, MDA-MB-231, and LLC1) were cultured as per the ATCC protocols. H1299 cell lines used in all our CM and coculture experiments were grown in MEM medium (Biological Industries). Medium was supplemented with 10% FBS (GIBCO, Thermo Fisher Scientific), 2 mmol/L l-glutamine, 1% Eagle nonessential amino acids, and Penicillin–Streptomycin solution (all from Biological Industries). Cancer cell lines authentication test was performed at the Genomics Center of Biomedical Core Facility, Technion. WM-266-4 and A-375 cells were a genreouse gift of Professor Gal Markel. To avoid the use of mycoplasma-contaminated cells, cells were routinely checked for mycoplasma using EZ-PCR Mycoplasma Detection Kit (Biological Industries).
Lung cancer CM
Lung cancer cell lines were cultured to 80% confluence in 10 cm dishes, washed twice with PBS, followed by adding MEM medium with no additives to the dish. Forty-eight hours later, the CM were collected and filtered using 0.22 μm Millex-GV filters (Merck Millipore). Control medium was prepared by parallel incubation of MEM medium without serum for 48 hours without cells (Fig. 1A). To prepare CM from H1299 pretreated with HK3T-CM, CM from HK3T cells was collected after 24 hours and transferred to H1299 cells for an additional 48 hours, than collected and filtered as described above (Fig. 1E). The control medium was either MEM incubated with no cells for 72 hours, or CM of HK3T cells. To prepare CM from ex vivo–cultured lung tissue specimens, tissues received directly from lung surgeries were utilized. Tumor or normal tissues were cut by a surgical blade, and tissue pieces measuring 0.3 to 0.5 cm3 were incubated in a 6 cm dish with 5 mL MEM medium with Penicillin–Streptomycin. The CM were collected after 48 hours and filtered as described above. To treat CAFs with CM, cells were washed twice with PBS, and CM was applied to the cells for incubation times indicated for each experiment.
Cycloheximide (C4859), MG-132 (M7449), and DMSO (D8418) were purchased from Sigma-Aldrich. Rapamycin (S1039), Torin-1 (S2827), and LY294002 (S1105) were procured from Selleckchem. Concentrations used in the experiments were 20 μmol/L cycloheximide, 25 μmol/L MG-132, or 0.1% DMSO (vol./vol.) as control was added to cancer-CM or control medium for 1 hour. In order to block translation of MDM2 after CM treatment, cycloheximide was added for 30 or 60 minutes after 1-hour treatment with either cancer-CM or control medium. Torin (0.2 μmol/L) was added to the cancer-CM or to the control medium for 1 hour. Rapamycin (1 μmol/L) and 10 μmol/L LY294002 were added to the growth medium 7 hours prior to the treatment with CM and for 1 hour during the following treatment with control or cancer-CM media.
Western blot analysis and antibodies
Cells were washed with ice-cold PBS, and protein was extracted in RIPA buffer (50 mmol/L Tris-HCl, pH 7.5, 150 mmol/L NaCl, 1% NP-40, 0.5% sodium deoxycholate, and 0.1% SDS). Total protein (40 μg) was separated on 7.5% SDS-polyacrylamide gel. Proteins were transferred to nitrocellulose membranes (Bio-Rad). Primary antibodies utilized for the detection of human MDM2 (mixture of 4B2, 2A9, and 4B11) and for human p53 (mixture of DOI and 1801) were a generous gift from Professor Moshe Oren. Antibodies for mouse MDM2 were obtained from Santa Cruz Biotechnology (SC-965). Anti-mTOR (catalog number #2983), anti-Akt (#4691), anti-p70 S6 Kinase (#9202), anti–phospho-mTOR Ser2448 (#5536), anti–phospho-Akt Ser473 (#4060), and anti–phospho-p70 S6 Kinase Thr389 (#9234) were obtained from Cell Signaling Technology. Anti-FN1 (HPA-027066), anti–beta-Actin (A-5441), and anti-Tubulin (T9026) were obtained from Sigma-Aldrich. Anti–α-SMA (ab5694) and anti-GAPDH (ab9485) were obtained from Abcam. Proteins were detected by horseradish peroxidase–conjugated secondary antibodies (Merck Millipore) followed by Luminata Crescendo Western HRP Substrate (Merck Millipore). For IHC staining, we used anti-human HLA-A (ab52922, Abcam) as per the manufacturer's recommendations.
Gene knockdown and overexpression
MDM2, p53, and AKT were knocked down using SMARTpool: siGENOME human MDM2 siRNA, human TP53 siRNA, and human AKT1 siRNA (Dharmacon). siGENOME nontargeting pool was used as control. The sequences of siRNA pools were as follows: si MDM2: Target Sequence 1: GCCAGUAUAUUAUGACUAA; Target Sequence 2: GAUGAGAAGCAACAACAUA; Target Sequence 3: CCCUAGGAAUUUAGACAAC; Target Sequence 4: AAAGUCUGUUGGUGCACAA. si TP53: Target Sequence 1: GAGGUUGGCUCUGACUGUA; Target Sequence 2: GCACAGAGGAAGAGAAUCU; Target Sequence 3: GAAGAAACCACUGGAUGGA; Target Sequence 4: GCUUCGAGAUGUUCCGAGA. si AKT1: Target Sequence 1:GACAAGGACGGGCACAUUA; Target Sequence 2: GCUACUUCCUCCUCAAGAA; Target Sequence 3: GACCGCCUCUGCUUUGUCA; Target Sequence 4: GGCAGCACGUGUACGAGAA. Nontargeting CONTROL: Sequence 1: UAAGGCUAUGAAGAGAUAC; Sequence 2: AUGUAUUGGCCUGUAUUAG. Sequence 3: AUGAACGUGAAUUGCUCAA; Target Sequence 4: UGGUUUACAUGUCGACUAA.
DharmaFECT reagent (Dharmacon) and siRNAs were used at final concentration of 50 nmol/L to transfect CAFs. Gene expression was tested 48 hours after transfection by Western blot analysis. Overexpression of human MDM2 was achieved by electroporation of pCMV-vector (empty or carrying the human MDM2 sequence, a kind gift from Professor Bert Vogelstein). CAFs were electroporated using Nucleofector 2b Device (Lonza) with AMAXA basic Nucleofector kit for primary fibroblasts (VPI-1002, LONZA).
To measure cell invasion, inserts (8 μm; FALCON, BD Biosciences) were coated with Matrigel (diluted 1:4 in serum-free MEM, BD Biosciences). H1299 cells (1.2 × 105 cells in 100 μL serum-free medium) were seeded on the insert. In a separate 24-well plate, CAFs were seeded a day before coculture (0.8 × 105 cells per well, in full medium). On the day of coculture, CAFs' medium was removed, cells were washed twice with PBS, and 600 μL of H1299-CM was added. As controls, wells without CAFs were used with either H1299-CM or a medium supplied with 10% FBS (as a positive control). The inserts containing H1299 were transferred to the CAFs plate for coculture. After 18 to 24 hours of incubation, all H1299 cells that remained in the insert were removed from the upper surface of the transwell membrane with a cotton swab. Invading cells on the lower membrane surface were fixed with 4% PFA, stained with Crystal violet, photographed, and quantified for total invasion area by relative arbitrary units using imageJ software, version 1.46r.
Quantitative real-time PCR
Total RNA was isolated using the PureLink RNA Mini Kit (Invitrogen) as per the manufacturer's protocol. RNA (2 μg) was reverse transcribed using a qScript cDNA synthesis kit (QuantaBio). QRT-PCR was performed using SYBR Green PCR Master Mix (Applied Biosystems) on a StepOnePlus instrument (Applied Biosystems). The expression level was normalized to that of the GAPDH housekeeping gene in the same sample. Primers were designed using Primer Express software. Primers for human MDM2—Forward primer: GTATCAGGCAGGGGAGAGTG; Reverse primer: GAAGCCAATTCTCACGAAGG. GAPDH—Forward primer: ACCCACTCCTCCACCTTTGA; Reverse primer: CTGTTGCTGTAGCCAAATTCGT.
In vivo tumor growth
H1299 cells expressing mCherry (1 × 106) with immortalized shp53 (2 × 106) CAFs electroporated with either empty plasmid or plasmid expressing human MDM2 were coinjected into the right flank of female nude mice (n = 10 per group; total inoculation volume = 50 μL). Tumor growth was visualized weekly using In Vivo Imaging System (IVIS) Spectrum (PerkinElmer, Inc.). Thirty days after injections, tumors were excised and fixed in 4% formalin for 48 hours to be further analyzed in IHC staining.
Comparison between invasion areas and tumor sizes was performed by the Student t test.
All patients whose specimens were included in the study provided written-informed consent for donating samples and data to the Sheba medical biorepository bank (#2019-SMC). The study was approved by the local ethics committee at Sheba Medical Center (#0226-13-SMC). Animal experiments were approved by Sheba Medical Center's Institutional Animal Care and Use Committee and by the Israeli Health Ministry (#985/15). Mice were monitored closely throughout all experimental protocols to minimize discomfort, distress, or pain. All methods were performed in accordance with the recommendations of the Panel on Euthanasia of the American Veterinary Medical Association.
Cancer-CM induces rapid elevation in stromal MDM2 protein
We have previously demonstrated that epithelial tumor cells can suppress DNA damage–induced p53 induction in neighboring fibroblasts, an effect mediated by tumor cell CM (45). While investigating p53′s repression mechanism, we found MDM2 protein to be rapidly induced in CAFs exposed to cancer-CM (Fig. 1). The elevation in MDM2 was transient, detected in lung CAFs after 1- to 2-hour exposure to lung cancer cells H1299-CM. No significant change was detected in p53 protein levels in the first few hours after exposure to cancer-CM; however, later on, p53 protein levels rose (possibly due to stress of serum starvation conditions), but this induction is observed similarly in CAFs treated with H1299-CM or with a control medium (Fig. 1B).
Longer exposure of CAFs to H1299-CM (up to 18 hours) did not induce any significant differences in MDM2 levels or in p53 in the absence of genotoxic stress (Supplementary Fig. S2). MDM2 protein elevation in CAFs' stromal cells was induced by H1299-CM, but not by lung fibroblasts-CM (Fig. 1C). Aiming to study the cancer cells' impact under conditions as close as possible to those in actual cancers, we used a freshly resected ex vivo–cultured lung tumor. Non–small cell lung carcinoma (NSCLC) was dissociated into small clumps immediately after surgical resection and incubated with medium for 48 hours to generate cancer-CM. As shown in Fig. 1D, the fresh lung tumor-CM strongly induced MDM2 elevation in CAFs. To further examine cancer-CM's effect on lung stromal fibroblasts, we produced primary, nonimmortalized fibroblasts isolated from either a normal part of the lung (NFs) or the tumor (CAFs) of lung adenocarcinomas and of lung squamous cell carcinomas (the two major NSCLC subtypes). The effect of cancer-CM on MDM2 elevation was similarly detected in all examined fibroblasts, whether CAFs or NFs, and regardless of histology of the tumor from which the fibroblasts were cultured (Supplementary Fig. S3). To evaluate reciprocal tumor–stroma cross-talk's role in our model, we treated HK3T CAFs with H1299-CM produced from H1299 cells preexposed to CM of HK3T (Fig. 1E). CM of H1299 cells pretreated with HK3T-CM induced MDM2 in the stroma to a greater extent than did CM of H1299 that were not pretreated with HK3T-CM (Fig. 1F), implying bidirectional cross-talk between cancer and stromal cells, mediated by secreted factor(s) and causing stromal MDM2 protein elevation.
MDM2 induction by cancer-CM is due to enhanced translation
MDM2 protein levels are modulated by transcriptional, posttranscriptional, and posttranslational mechanisms (48, 49). We found that mRNA levels of MDM2 in fibroblasts were not affected by cancer-CM in the relevant time frame (Fig. 2A). In p53-knocked down fibroblasts, MDM2 protein levels are barely detectable (demonstrated later on in Fig. 5), consistent with p53 being the major regulator of MDM2 transcription (24). The borderline visualization of MDM2 by Western blots in those cells precludes us from reaching a concrete conclusion about the impact of cancer-CM on MDM2 in the absence of p53 (to be further discussed below). We next examined the possibility that reduced degradation of MDM2 protein in the presence of cancer-CM is the cause for its higher levels. As shown in Fig. 2B and C, blocking translation by cycloheximide after 1 hour of MDM2 induction by cancer-CM demonstrates equally rapid degradation of MDM2 protein in the presence of control or cancer-CM.
To further evaluate MDM2 degradation in the presence of cancer-CM, we blocked protein proteasomal degradation using a proteasomal inhibitor (MG-132; Fig. 2D). Although MG-132 treatment resulted in accumulation of MDM2 protein as expected, it did not eliminate the difference between cancer-CM and control-treated CAFs. We thus conclude that reduced MDM2 protein degradation is not the reason for higher MDM2 protein levels in CAFs treated with cancer-CM.
We next tested the impact of addition of cycloheximide from the beginning of the CM treatment; under these conditions, MDM2 protein levels were nondetectable with or without the addition of cancer-CM (Fig. 2E). These results demonstrate that MDM2 undergoes rapid turnover in CAFs. They further suggest that cancer-CM enhances stromal-MDM2 protein levels through enhanced translation.
MDM2 elevation is mediated by mTORC1 and correlates with AKT and S6K1 phosphorylation
PI3K and AKT pathways have been reported as modulating MDM2 localization and activity (50). Specifically, downstream to PI3K, mTORC1 regulates cap-dependent protein translation through phosphorylation of eukaryotic initiation factor 4E–binding proteins, ribosomal protein S6 kinases (S6K), eIF4G, La Ribonucleoprotein Domain Family Member 1, and repressor of Pol III transcription MAF1 homologue (MAF1; ref. 51). Importantly, mTOR has been demonstrated to enhance MDM2 protein translation (52–54). In addition, downstream to mTOR, p70 S6 kinase 1 (also called S6K-B1; S6K1) binds, phosphorylates, and inhibits MDM2 activity as a p53 E3-ligase (55). Using PI3K/mTOR inhibitor (LY294002) and mTOR inhibitors (Rapamycin and Torin-1), we demonstrated the MDM2-inducing effect of cancer-CM to be dependent upon both PI3K/AKT pathway and mTOR activity (Fig. 3A and B). Rapamycin's efficacy in preventing stromal-MDM2 elevation points to mTORC1—and not mTORC2—as mediating this effect (56). Interestingly, although rapamycin as well as LY294002 prevented cancer-CM from inducing MDM2, each of these inhibitors elevated basal MDM2 levels on its own (Fig. 3A), and a similar effect was seen with torin treatment (Fig. 3B). Several negative feedback mechanisms are known to inhibit PI3K/AKT following mTORC1 activation, including mTORC1-mediated growth factor receptor-bound protein 10 (GRB10) phosphorylation and insulin receptor substrates 1/2 degradation (57, 58). Therefore, the elevation of MDM2 levels by mTOR inhibition suggests an additional, mTORC1-independent mechanism whereby the PI3K/AKT signaling elevates MDM2 levels. Extensive cross-talk exists between this signaling cascade and other growth factor signaling pathways and can potentially explain this observation (59).
To clarify whether activation of AKT is involved in stromal-MDM2 induction, we evaluated AKT phosphorylation in cancer-CM–treated CAFs. Strong correlation between mTOR phosphorylation (Serine 2448), AKT phosphorylation (Serine 473), and MDM2 protein levels was detected (Fig. 3B). In order to further test if the elevation of MDM2 is controlled by CM-induced AKT phosphorylation, we abrogated Akt1 expression in CAFs by targeted siRNA and treated them with H1299-CM for 1 hour. As shown in Fig. 3C, knocking down Akt in CAFs prevented the CM-induced MDM2 elevation. Of note, CM-induced phosphorylation of mTOR was also reduced in AKT knocked down cells. These results support our conclusion that cancer-CM induction of AKT activity, followed by activation of mTORC1, is required for MDM2 elevation (56). Examination of the mTORC1 translation-related downstream signaling pathway revealed that phosphorylation of p70 S6 Kinase (Thr389), a downstream target of mTORC1, is elevated in stromal cells by cancer-CMs, similarly to MDM2 (Fig. 3D; Supplementary Fig. S4). These results suggest involvement of AKT, mTORC1, and p70 S6 Kinase activation in stromal cells' response to cancer-CM, bringing about enhanced stromal-MDM2 translation.
Robustness of stromal-MDM2 protein induction
To examine whether MDM2 protein induction by cancer cells is a generalized phenomenon, we tested CM produced by a variety of cancer cell lines for its impact on stromal-MDM2. As demonstrated (Figs. 3D and 4), stromal-MDM2 induction is seen following exposure to a variety of cancer cells-CM including lung cancer, breast cancer, and melanoma cell lines, in contrast to minimal or no induction by CM of non-cancerous fibroblasts (grown from resected specimens of several patients with lung cancer). MDM2 protein induction was found to closely correlate with enhanced levels of phosphorylated (ser2448) mTOR. We conclude that cancer-CM robustly enhances MDM2 translation and its protein levels in stromal fibroblasts, an effect mediated by mTOR signaling.
Stromal mTOR signaling and MDM2 protein were variably induced by CM of NSCLC freshly resected, ex vivo–cultured tumor CM in comparison with CM of similarly cultured normal lung specimens from the same patient (Supplementary Fig. S4, Patient #2). Interestingly, CM from lung carcinoid (a low-grade neuroendocrine cancer with a relatively low risk of metastatic spread) induced the stromal mTOR-MDM2 pathway to a lesser extent (Patient #1, Supplementary Fig. S4) than did CM originated from the adenocarcinoma NSCLC tumor (Patient #2). The clinical relevance of this phenomenon requires a larger study.
As additional evidence of stromal-MDM2 induction by cancer-CM's robustness, mouse cancer-CM induced elevation of MDM2 protein levels in mouse primary fibroblasts (Supplementary Fig. S5). Mouse cancer-CM also markedly activated mouse mTOR in mouse primary fibroblasts, but had no impact on human-stromal-MDM2 (Supplementary Fig. S5). In contrast, human cancer-CM did induce mouse fibroblasts-MDM2, albeit to a lesser extent than mouse cancer-CM's effect (Supplementary Fig. S5), demonstrating that the effect is mostly, although not completely, species specific.
Stromal-MDM2 and AKT are required for a proinvasive effect that is p53-independent
CAFs' procancerous role in TME has been reported in the past, specifically demonstrated to enhance invasiveness in neighboring cancer cells (60). We wanted to explore the possibility that CAFs' mTOR-MDM2 signaling modulates enhanced invasiveness of neighboring cancer cells. CAFs were knocked down for MDM2 (Supplementary Fig. S6) resulting in attenuation of H1299 cancer cell invasion toward them. Because MDM2 is a negative regulator of p53, we explored the possibility that siMDM2′s effect is mediated through the expected upregulation of stromal-p53. We took advantage of HK3T CAFs stably expressing shp53 plasmid, resulting in p53 gene knockdown (Fig. 5A), and compared it with control cells expressing shmNOXA (an irrelevant mouse NOXA knockdown sequence). shp53 cells harbored extremely low basal levels of MDM2, as expected in cells lacking p53 (Fig. 5A). Indeed, no MDM2 elevation was detected in these cells after H1299-CM treatment. Importantly, activation of the mTOR-AKT pathway was induced is those cells after exposure to cancer-CM regardless p53 status (Fig. 5B). Next, we knocked down MDM2 in both cell types and tested them as invasion inducers when cocultured with H1299 (Fig. 5C and D). Knocking down MDM2 in control cells resulted in reduction of H1299 cancer cell invasion, whereas HK3T shp53 cells did not induce cancer cell invasion, irrespective of MDM2 knockdown (Fig. 5C–E). Cancer-CM's inhibitory effect on p53 expression that we reported on previously (45) is unlikely to play a role here, as it is seen only under conditions of DNA damage signaling. Moreover, according to a recently published report, stromal-p53 in lung cancer CAFs can have a procancer and proinvasive influence on neighboring cancer cells (12); this is consistent with our observation of shp53 cells having no invasion-enhancing effect (see also Supplementary Fig. S7) and is in contrast to what could be expected based on previous reports (43, 61–63). Of note, addition of mTOR inhibitor rapamycin to the medium reduced H1299′s invasiveness regardless of the stromal component, as its effect was also apparent on H1299 alone (Supplementary Fig. S6). We conclude that CAFs' paracrine invasion-promoting effect is largely dependent upon stromal MDM2.
Next, we utilized Akt-1 knocked down CAFs to query whether the level of AKT activation will affect the invasion of H1299 cocultured cells. Indeed, Akt-1 knocked down CAFs failed to induce invasion of H1299 cells (Fig. 5F–H). This result corroborates the fact that MDM2 levels are not elevated by cancer-CM in Akt-1 knocked down cells (Fig. 3C), and further supports the role of stromal-AKT in the tumor–host cross-talk we delineate here, and in promoting invasion of neighboring tumor cells.
Overexpression of MDM2 in stromal cells enhances cancer invasion
To further elucidate stromal-p53′s role in the cancer cell invasion induced by stromal-MDM2, we transfected human MDM2 into HK3T shp53 cells to test whether their proinvasive influence would be recovered. Indeed, MDM2 overexpression in these cells led to a significant elevated cancer cell invasion (Fig. 6; Supplementary Fig. S7), supporting the notion that stromal-MDM2 has direct impact on neighboring cancer cell invasion in a p53-independent manner. Of note, overexpression of MDM2 in control cells (shmNOXA) did not enhance their proinvasive properties (Supplementary Fig. S7). This result might indicate that in these cells, the endogenous MDM2 (induced by the cocultured cancer cells) activates maximally the proinvasive signaling to the neighboring cancer cells, unlike in shp53 cells wherein endogenous MDM2 levels are markedly attenuated by p53 knockdown.
Stromal-MDM2 induces tumor growth in vivo
To examine stromal-MDM2′s role on tumor growth in vivo, we coinoculated in nude mice H1299 cells (expressing mCherry) together with immortalized shp53 HK3T CAFs overexpressing either a human MDM2 expression plasmid or a control plasmid. Tumor size was evaluated using IVIS. Thirty days after injection, tumors were excised and analyzed. Despite variability in tumor growth, tumors containing MDM2-expressing fibroblasts were significantly larger (Fig. 7A). Pathologic examination of the inoculated areas revealed even more dramatic differences, with mostly degraded tumor areas or nondetectable tumor cells in the control mice, whereas mice injected with MDM2-overexpressing CAFs exhibited large, dense epithelial tumors (Fig. 7B). Staining with α-human HLA-A confirmed the tumor cells to be of human origin and that the tumors are of epithelial cell origin as expected and not fibroblasts (Supplementary Fig. S8). These results suggest that MDM2 expression in the stroma supports tumor growth in vivo. As expected and as demonstrated by the histologic evaluation of the tumors, the inoculated fibroblasts do not persist in the tumor. This being the case, alongside the fact that overexpression of MDM2 in the fibroblasts is transient, we conclude the impact of stromal-MDM2 to be critical in the initial, establishment phase of the tumors in vivo.
We have identified a novel tumor–host bidirectional signaling, involving stromal MDM2 induction by cancer-originated soluble factor(s), and reciprocal paracrine stromal MDM2-dependent invasion-promoting impact of the stroma. This interaction appears to have a positive feedback nature, as cancer cells exposed to stroma-CM induce stromal-MDM2 better than do cancer cells not previously exposed to stroma-CM (Fig. 1E and F). Stromal-MDM2 is induced by malignant cells' CM, and not induced or only minimally induced by CM of nontransformed cells such as primary fibroblasts (Fig. 4). Furthermore, a variety of cancer cell lines demonstrated a stromal-MDM2–inducing effect, as did specimens of human lung cancers cultured fresh ex vivo. Interestingly, a similar stromal-mTOR/MDM2 induction effect was seen in mouse fibroblasts exposed to mouse-tumor-CM (Supplementary Fig. S5).
Although our data suggest a correlation between tumor aggressiveness and its ability to induce stromal-MDM2 (Supplementary Fig. S4), this hypothesis requires further investigation. Importantly, stromal-MDM2 knockdown reduced neighboring cancer cells' invasiveness (Fig. 5), while overexpression of stromal-MDM2 enhanced this invasion (Fig. 6), as well as promoted cancer growth in vivo (Fig. 7). Therefore, this novel tumor–host interaction may have important clinical implications for a variety of cancers, and understanding its molecular mediators may provide novel therapeutic targets.
A variety of signals and mechanisms can regulate MDM2 protein levels, reflecting MDM2′s importance as a regulator of critical cell-fate choices. The most recognized MDM2 regulator is p53, which activates the gene's transcription. However, cancer-CM did not induce MDM2 mRNA levels, indicating this phenomenon to be most likely p53-independent. We found AKT to be strongly phosphorylated in stromal cells immediately following exposure to cancer-CM, both in p53-deficient or proficient cells. The increase in stromal-MDM2 levels was dependent upon AKT expression, mTORC1 activation, and to be accompanied by elevation of phosphorylated AKT and p70 S6 Kinase. Moreover, we provide evidence of enhanced MDM2 translation under these conditions (Fig. 2). We conclude therefore that mTORC1 complex is activated by cancer-CM, in turn activating translation enhancement signaling that promotes MDM2 translation and protein accumulation. From the lack of significant MDM2 cross-induction between mouse and human cells (Supplementary Fig. S5), we conclude that a specific signal—and not general lack of nutrients in the CM—is involved in MDM2 induction by cancer-CM. The identity of the mediator remains to be discovered and is not the focus of this article.
MDM2 has been shown to enhance a wide variety of procancerous signals and processes besides p53 downregulation. Although initially believed to have a role in the p53-MDM2 negative feedback loop only, mounting evidence points to roles in stemness, redox homeostasis, EMT, angiogenesis, invasion, and metastatic spread (29, 30, 33, 40, 64). Although the molecular mechanisms involved vary, they plausibly involve chromatin modulation through its histone ubiquitination activity, or other protein–protein interactions (34). Although many possible MDM2 procancer effects have been described, to our knowledge no studies to date have studied stromal-MDM2′s effect on its microenvironment. We demonstrate for the first time that stromal-MDM2 activation can promote invasion of neighboring cancer cells. In vivo, the impact of stromal-MDM2 seems to occur during the tumor establishment phase. Notably, some of the work demonstrating MDM2′s role in promoting stemness was performed in mouse embryonic fibroblasts (33), thus the tumor-CM causing elevation of stromal-MDM2 can be speculated to enhance the CAFs' stemness and replicative potential, along with the potential of promoting cancer growth as a result of CAF proliferation.
The initial hypothesis and observations that led to this study were related to the inhibitory impact of stromal-p53 on neighboring cancer cells (45, 46). We reported in the past that cancer-CM attenuates stromal-p53 induction by DNA damage. Although the effect on p53 is seen only under conditions of genotoxic stress, and at later time points than those of MDM2 induction, we cannot rule out that the induction of stromal-MDM2 that we report now is related to the p53-suppression effect; MDM2 might bind p53 following MDM2′s initial early induction or otherwise modify it, thereby preventing a later p53 induction by DNA damage. MDM2-p53 binding was examined in our studies; no difference dependent upon cancer-CM presence could be detected (data not shown).
A recent study demonstrating a wild-type stromal-p53 protumor influence, specifically in CAFs of lung cancer (ref. 12; similar to the cells utilized in our study), complicates our understanding of the potential effect of stromal-MDM2 induction on the p53 pathway, and on its eventual influence on neighboring tumor cells. It can be speculated that stromal-MDM2 induction participates in redirecting wild-type p53 activity within CAFs to a tumor-supportive role. This hypothesis was not examined thus far. However, the invasion-promoting role of stromal-MDM2 we describe herein is independent of stromal-p53. Further studies are required to elucidate the impact of the cancer-induced stromal-MDM2 on stromal-p53 activity and their compound influence on cancer initiation, invasion, and spread.
Disclosure of Potential Conflicts of Interest
A. Onn has received honoraria from the speakers' bureau of Boehringer Ingelheim and AstraZeneca. J. Bar reports receiving commercial research grants from MSD, AstraZeneca, and Pfizer. No potential conflicts of interest were disclosed by the other authors.
Conception and design: I. Kamer, I. Daniel-Meshulam, A. Onn, J. Bar
Development of methodology: I. Kamer, M. Perelman, J. Bar
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): I. Kamer, I. Daniel-Meshulam, O. Zadok, E. Bab-Dinitz, G. Perry, R. Feniger-Barish, M. Perelman, I. Barshack, A. Ben-Nun, J. Bar
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): I. Kamer, I. Daniel-Meshulam, I. Barshack, J. Bar
Writing, review, and/or revision of the manuscript: I. Kamer, I. Daniel-Meshulam, O. Zadok, E. Bab-Dinitz, G. Perry, R. Feniger-Barish, M. Perelman, I. Barshack, A. Ben-Nun, A. Onn, J. Bar
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): I. Kamer, I. Barshack, J. Bar
Study supervision: A. Ben-Nun, J. Bar
This work was supported by the Israel Science Foundation (grant no. 0333100; supported I. Kamer and G. Perry), by the Israel Cancer Association through ICA-USA Board of Directors (grant no. 20160126; supported O. Zadok), and by the late Jannette Horowitz's estate (supported I. Danial-Meshulam).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.