Early-Stage Metastasis Requires Mdm2 and Not p53 Gain of Function

Metastatic foci arise after primary tumor cells successfully transition through various phases of the metastatic process. These phases include invasive migration from the primary tumor, intravasation into the blood stream, extravasation into a distant organ, and colonization in conducive tissues. In response to avascular hypoxic conditions, primary tumor cells may produce VEGF, which stimulates angiogenesis, recruits endothelial cells for neovascularization, and/or permeabilizes existing vasculature through which tumor cells can intravasate. The dissemination of tumor cells to distant organs can be initiated through an intracellular program called epithelial–mesenchymal-transition (EMT). The programming of tumor cells to migrate requires the upregulation of transcription factors, such as Snail Family Zinc Finger 1 (SNAI1), and increased levels of adhesion proteins and filaments, including N-cadherin and vimentin. Conversely, E-cadherin is decreased as cells transition into a mesenchymal phenotype. Moreover, we have shown that the oncogene Mouse double minute 2 (Mdm2) is also positively correlated with EMT (1–5). Clinically, mdm2 gene amplification occurs in approximately 10% of all human cancers, yet more importantly, several studies have detected the Mdm2 protein in 40% to 80% of high-grade human tumors (6, 7). Increased levels of Mdm2 are correlated with poor prognosis, especially in breast cancer (8), and several correlative studies have inferred its involvement in metastasis (9–12). Similarly, we have previously shown that Mdm2 is elevated in 73% of breast carcinomas with known metastases (1).

Although several reports have asserted that Mdm2 promotes metastasis, they were either correlative studies using human patient samples (9, 10, 12), or in vitro studies with or without a correlative component (1, 3–5). Only two studies thus far have explored the importance of Mdm2 in metastasis in vivo. One study, conducted in pancreatic carcinoma cells, found that silencing Mdm2 resulted in decreased proliferation (2). This study, along with another that used tail vein injection of transiently transfected KHT-C murine cells (11), employed approaches that did not directly assess the initiation or other phases of metastasis to define the role of Mdm2. The authors of both studies attributed the differences in metastasis to the ability of Mdm2 to downregulate wild-type p53. Early reports show that Mdm2 can facilitate the destabilization of p53 in response to genotoxic stress; however, 20% to 40% of breast cancers subtypes (13), and 80% of basal-like breast cancers (14) maintain mutant p53 (mutp53). Considering the gain-of-function properties of mutant p53, it might be expected that any downregulation of mutant p53 would result in decreased metastasis. Regardless, unlike p53, mutp53 is unable to induce mdm2 gene expression. We recently demonstrated that the TGFβ1–Smad3/4 axis transcriptionally activates the mdm2 gene independently of p53. Furthermore, the activated TGFβ1 pathway and elevated Mdm2 levels were correlated with human invasive/metastatic breast cancer specimens (1). Here, we show that Mdm2 is necessary for upregulating essential markers of EMT, and consequently, migration/invasion. In addition, we show that Mdm2 is important in upregulating MMP-2 (matrix metalloproteinase 2), an indicator of highly metastatic cancers that colonize the lung. Our orthotopic experiments show that Mdm2 is a key mediator of metastatic lung foci through regulation of angiogenesis and intravasation. These events are independent of mutp53 and demonstrate a distinct role for Mdm2 at specific stages of the metastatic process.

shMdm2 was obtained in pLKO from Open Biosystems (TRCN0000003380, which targets exon 12). Additional cell lines expressing shMdm2 constructs (TRCN0000003376 and TRCN0000003377) were also examined to confirm the phenotypes were due to knockdown of Mdm2. Previously described vectors include shp53 (15, 16), MCHERRY (17), and EGFP (17). Lentivirus was produced as described previously (16). Briefly, Lentiviral vectors were packaged in 293T cells using second-generation packaging constructs: pCMV- dR8.74 and pMD2G. Supernatant media containing virus were collected at 36 to 48 hours, supplemented with 4 μg/mL polybrene, filtered through a 0.45-μm filter, and added to MDA-MB-468 (R273H p53) and TMD231 cells [a derivative of MDA-MB-231 cells, which metastasize to the lung (18) and have R280K p53]. Cells were selected for puromycin (2 μg/mL) resistance. All cells were grown as described previously (1). Cells were incubated under normoxia (21% oxygen in complete DMEM), hypoxia (1% oxygen in complete DMEM/F-12 1:1 HyClone), and/or with 10 ng/mL TGFβ1 for the times indicated. Western blot analysis was performed as described by Waning and colleagues (19) using antibodies: Mdm2 [IF2 (Ab-1; OP46), and 4B2 (OP145) from Calbiochem and 2A10 (ab16895) from Abcam; N-cadherin (610920) from BDBioSci]; p53 (DO-1; sc-126), GAPDH (6C5; sc-32233), vinculin (N-19; sc-7649 or E1E9V-13901p from Cell Signaling Technology), E-cadherin (G-10; sc-8426 and 67A4; sc-2179), vimentin (RV202; sc-32322), VEGF (147; sc-507), tubulin (TU-02; sc-8035), and PAI (H-135-sc-8979) from Santa Cruz Biotechnology; Snail (F.31.8- MA5-14801 from Thermo Fisher Scientific); actin (A1978) from Sigma; HIF1α (NBP1-02160) from Novus; and MMP2 (MAB-3308) from Millipore.

All cells were tested for mycoplasma with the Mycoplasma Detection Kit (InvivoGen) during the experiments and before injection into animals; TMD231 cells have a mesenchymal morphology and MDA-MB-468 have an epithelial-like morphology, which was verified by microscopy. Molecularly, TMD231 do not express E-cadherin, and MDA-MB-468 express E-cadherin, which is verified by Western blot analysis. The TMD231 cells were generated in-house from MDA-MB-231 cells by Dr. Nakshatri, Indiana University School of Medicine, Indianapolis. MDA-MB-468 cells were obtained from ATCC. All cells were passaged for no more than 3 months between thawing and use in experiments.

Migration of MDA468 cells was assessed using culture inserts (IBIDI) according to the manufacturer's instructions. Briefly, 2.45 × 10⁴ MDA468 shControl cells and 4.9 × 10⁴ MDA468 shMdm2 cells were placed into both wells of the culture insert (separate insert for each cell line). After a 48-hour incubation, inserts were removed, and images were obtained at the indicated time points. Gap area was quantified using the wound healing tool for imageJ (http://dev.mri.cnrs.fr/projects/imagej-macros/wiki/Wound_Healing_Tool). Inserts were manufactured to result in a 500-μm gap. Time 0 hour was imaged at ×5 magnification, and all subsequent time points were imaged at ×10 magnification.

TMD231 mCherry shControl and GFP shMdm2 cells were plated onto coverslips in a 35-mm plate (64,000 cells/plate). Coverslips were then transferred to a 60-mm plate and secured with paramount the next day. Images were obtained immediately and at the times indicated; the number of cells of each color in three separate fields of view were counted and averaged.

Boyden chamber assays were conducted with BD Matrigel Invasion chamber 8.0-μm pore size according to the manufacturer's instructions. Briefly, 2.5 × 10⁴ cells were seeded into each well of Matrigel chambers, with 5% serum used as the chemoattractant. After a 22-hour incubation under normal growth conditions, the media were removed, and the surface of the membrane was wiped with a cotton tip swab to remove noninvading cells. Membranes were then removed from the chambers and stained by incubating in methanol for 2 minutes and then 1% toluidine blue for 2 minutes. Cells were counted in three random fields at a magnification of ×20, and the experiment was done in triplicate. Percent invasion was calculated by dividing the mean number of cells invading through the Matrigel insert membrane by the mean number of cells migrating through the control insert membrane.

All animal studies have been approved by Indiana University School of Medicine IACUC (protocol #11190) and were performed by the Indiana University In Vivo Therapeutics Core. A total of 7.5 × 10⁵ TMD 231 shGFP, or shMdm2 cells or 1 × 10⁶ MDA468 shControl or shMdm2 cells were injected into the fourth mammary fat pad of NOD/SCID gamma mice (n = 7/group) obtained from the In Vivo Therapeutics Core of the Indiana University Simon Cancer Center (Indianapolis, IN). Tumor growth was calculated by volumetric analysis using calipers. Tumors were allowed to grow to an average volume of 600 mm³ (TMD231 and MDA468 shControl), 200 mm³ (MDA468 shMdm2), or 900 mm³ (TMD231 in the shp53 experiment).

Cardiac punctures with 1 mL TB syringe and a 27 g × 0.5 needle (BD Biosciences) were performed at necropsy and temporarily stored in vacutainers coated with sodium heparin (BD Biosciences) prior to peripheral blood mononuclear cells (PBMC) processing. PBMCs were isolated by Ficoll–Paque Plus (GE Healthcare) separation and cells were stained with 1 μg PE-conjugated mouse anti-human EGF receptor antibody (#555997; BD Pharmigen) per 1 × 10⁶ cells. FACS analysis was conducted and data analysis was performed using FlowJo 10.0.8r1. The percentage of circulating tumor cell (CTC) population of each shControl and shMdm2 sample was determined by positive human-specific EGFR staining of CTCs in mouse blood.

Tissues were fixed overnight at room temperature in 10% neutral-buffered formalin, after which they were transferred through graded concentrations of alcohol to xylene inside a Leica Automated Vacuum Tissue Processor. Samples were embedded in paraffin before being cut into 5-μm thick sections, mounted onto positively charged slides, and baked at 60°C.

The Indiana University Pathology Core performed all tumor preparation, hematoxylin and eosin (H&E) and CD31 staining, and quantitation. Slides were then deparaffinized in xylene and rehydrated through graded alcohols to water. Antigen retrieval was performed by immersing the slides in Target Retrieval Solution (Dako) for 20 minutes at 90°C, cooling at room temperature for 10 minutes, washing in water, and then proceeding with immunostaining. Slides were blocked with protein blocking solution (Dako) for 30 minutes. All subsequent staining steps were performed using the Dako FLEX SYSTEM on an automated Immunostainer; incubations were done at room temperature and Tris buffered saline plus 0.05% Tween 20, pH 7.4 (TBS; Dako Corp.) was used for all washes and diluents. Thorough washing was performed after each incubation. Standard H&E staining was then performed. Alternatively, slides were stained with anti-mouse CD31 (DIA 310; Dianova). Control sections were treated with an isotype control using the same concentration as primary antibodies to verify the staining specificity.

Aperio's whole slide digital imaging system, ScanScope CS, was used for imaging. All slides were imaged at 20×. Digital images were obtained using Aperio's software, ImageScope. For CD31 analysis, whole primary tumor cross-sections were analyzed, with the exception of the excluded necrotic tissue. Vessels less than 125 μm and/or having a diameter less than 30 μm were considered immature. Data from 5 independent hand counts per tumor were confirmed by a pathologist, averaged, and presented as the average percent of immature vessels (the average number of immature vessels divided by the average total vessels in each primary tumor).

H&E evaluation of lung metastases was performed by a pathologic assistant. All hand counts matched pathological reads performed by a board-certified pathologist. Metastases from all lungs were classified as small (1–25 cells), medium (26–100 cells), large (101–500 cells), or very large (501+). Represented values are averages from each corresponding metastatic size group.

The entire lung of each mouse was scored for metastases and divided by the area of each lung.

To confirm that Mdm2 is required for metastasis, MDA468 and TMD231 cells were transduced with control and Mdm2-specific shRNA lentiviruses. MDA468 cells, a basal-like cell line, can undergo EMT while TMD231 cells, a derivative of MDA231 cells, are post-EMT and resemble mesenchymal cells. Once the cell lines were confirmed for decreased levels of Mdm2 (Fig. 1A; Supplementary Fig. S1), we analyzed levels of mutp53 as Mdm2 can regulate p53, and many studies have demonstrated that mutp53 promotes tumorigenesis and metastasis (20–38). Interestingly, knockdown of Mdm2 in the MDA468 cells resulted in an increase in mutp53 levels, while no discernable difference was observed when TMD231 shMdm2 and shControl cells were compared (Fig. 1A). We next explored whether shMdm2 cell lines would undergo EMT. The mesenchymal marker, SNAI1, was decreased in both shMdm2 cell lines (Fig. 1B). EMT markers in shControl and shMdm2 MDA468 cells were further examined, and shMdm2 cells were found to molecularly resemble epithelial cells with regards to E-cadherin, N-cadherin, and vimentin (Fig. 1C; Supplementary Fig. S1). Silencing Mdm2 also resulted in a more epithelial-like morphology compared with the shControl cells (Fig. 1D), which is consistent with the low level of EMT markers. MDA468 cells were treated with TGFβ1 and hypoxia to induce EMT. While TGFβ1 treatment of shControl cells resulted in increased levels of various mesenchymal markers and a more mesenchymal morphology, it did not affect the markers or alter the epithelial-like morphology of shMdm2 cells. TGFβ1 treatment of TMD231 cells did not cause EMT, as these cells are morphologically (spindle-shaped) and molecularly (no E-cadherin, high snail, vimentin, N-cadherin) mesenchymal like (Supplementary Fig. S1). Interestingly, despite having increased mutp53 levels, shMdm2 MDA468 cells did not have a mesenchymal phenotype, suggesting a mutant p53-independent role for Mdm2 in EMT. Taken together, these results indicate that Mdm2 plays an important role in regulating key intracellular programing for EMT.

As migration is coupled with EMT, we tested whether Mdm2 affected cellular migration. Initially, we conducted traditional wound healing assays and quantified gap closure (Fig. 2A). Gap closure was dramatically delayed in shMdm2 MDA468 cells (50% at 70 hours) compared with shControl cells (completed at 40 hours). To confirm these data, we also conducted a slight variation of the wound healing assay by examining the migration of fluorescent cells off of coverslips. A GFP-shMdm2 plated coverslip and a mCherry shControl plated coverslip were transferred and fixed on a new plate. Cells that migrated into the void between the coverslips were then counted. This latter technique allowed us to compare shMdm2 and shControl cells under the same conditions. shControl cells were more effective at migration than shMdm2 cells (Fig. 2B). Finally, using a Boyden chamber assay, we demonstrated that shControl cells were able to invade through Matrigel more effectively than shMdm2 cells (Fig. 2C). Thus, using three independent methods, our results show that migration and invasion are diminished with decreased Mdm2.

As cell growth and migration can be interrelated, and shMdm2 cells exhibit a defect in migration, we performed cell-cycle analysis by FACS. Interestingly, although both cell lines have mutp53 (elevated in shMdm2 MDA468), there were no differences in cell-cycle distribution between shControl and shMdm2 (Supplementary Table S1). These data eliminate the role of Mdm2 in regulating gain-of-function mutp53 in cell cycle and migration. To garner insight into the role of Mdm2 in in vivo tumor growth, we orthotopically injected shControl or shMdm2 MDA468 or TMD231 into the mammary fat pad of 6-week-old NSG female mice (n = 7 for each group). The average tumor volumes and weights of shMdm2 MDA468 tumors were lower than shControl tumors (Fig. 3A and B, left), despite allowing them to develop for 30 days longer. This result is notable because mutp53 is thought to enhance tumorigenicity, and its levels are higher in the shMdm2 MDA468 cell line. There were no significant differences between the final TMD231 tumor volumes or weight at the time of euthanasia (Fig. 3A and B, right).

Considering the delay in tumor growth in both TMD231 and MDA468 shMdm2 cells, we examined in vitro levels of key factors associated with angiogenesis. We and others have shown that Mdm2 can play a role in stabilizing hypoxia-inducible factor 1α (HIF1α) and upregulating VEGF (39–41). Mdm2 can also stabilize VEGF mRNA levels to promote angiogenesis (42). In accordance with these published findings, there was a substantial decrease in the levels of HIF1α and VEGF in both the cultured TMD231 and MDA468 lines grown under normoxia or hypoxia when Mdm2 was knocked down compared with the shControl (Fig. 4A). On the basis of these in vitro results, primary tumors were sectioned and the endothelial cells were stained with cluster of differentiation 31 (CD31). In the presence of Mdm2, the percentage of newly formed, immature vessels in both MDA468 and TMD231 shControl tumors was significantly higher than shMdm2 tumors (Fig. 4B). We next analyzed lysates from MDA468 and TMD231 shMdm2 tumor sets to confirm the in vitro results. In agreement with the levels seen in cell culture (Fig. 4A), Figure 4C demonstrates diminished HIF1α and VEGF levels in shMdm2 tumors compared with shControl tumors. Taken together, our in vitro and in vivo data are congruent in showing that Mdm2 is integrated into the angiogenesis pathway.

Although our data suggest Mdm2 is an important driver of angiogenesis, previous reports implicate mutp53 is also important for this phenomenon (27, 43–45). Considering that our cells maintain mutp53, we tested the importance of this protein in angiogenesis by implanting TMD231 cells that had been transduced with either shControl or shp53 plasmid (Supplementary Fig. S2). Surprisingly, shp53 tumors had a significantly higher percentage of immature vessels compared with shControl tumors (Fig. 4B). Analysis of tumor extracts revealed that Mdm2 levels were slightly elevated in shp53 tumors (Supplementary Fig. S2). Thus, Mdm2 promotes vascularization of the tumor via regulation of the HIF1α–VEGF axis and possibly permits intravasation of tumor cells into the blood stream.

Tumor cells undergo intracellular reprograming to signal for cellular migration. To assess whether these changes were evident in our in vivo metastasis models, we examined EMT markers from shMdm2 MDA468 tumor extracts. The results support our in vitro findings of decreased levels of N-cadherin and SNAI1 and increased levels of E-cadherin compared with the shControl tumors (Fig. 5A). Two additional proteins that are associated with metastasis, plasminogen activator inhibitor-1 (PAI-1), an important tumor angiogenic factor involved in metastasis (46), and MMP2, a protein that acts to break down extracellular matrix and is associated with metastatic lung foci (47) were examined by Western blot analysis of tumor extracts. Both PAI-1 and MMP2 were found to be decreased in shMdm2 primary tumors (Fig. 5A).

These data indicate that the tumor cells are primed to intravasate and, with our angiogenesis data in Fig. 4, suggest that shControl tumor cells would enter into the blood stream. We extracted blood from the euthanized animals harboring shControl and shMdm2 tumors and stained for human EGFR to detect CTCs by FACS analysis. We found a greater than 4-fold increase in CTC in animals implanted with MDA468 shControl cells compared with those implanted with the shMdm2 cells (Fig. 5B). Conversely, a greater number of shp53 CTC versus shControl CTC was observed. This result is in accordance with the vasculature data shown in Fig. 4, but contradicts published findings that suggest mutant p53 is driving metastasis (20–23, 25–37, 48–50). Thus, in the presence of Mdm2, tumor cells are priming the vasculature and entering the blood stream.

After establishing that the numbers of tumor cells entering the blood stream were dependent on Mdm2, we determined whether a corresponding change in lung metastatic foci was evident. MDA468 shControl cells readily metastasized to the lung with an average of 1 foci/mm², whereas the shMdm2 cells averaged only 0.15 foci in the same area. When Mdm2 was silenced, an 85% decrease in lung foci was observed even with elevated mutp53 (Fig. 5C, top). Similarly, the TMD231 shControl cells were significantly more likely to metastasize to the lung than shMdm2 cells (Fig. 5C, middle). Surprisingly, considering the higher number of CTCs in shp53 TMD231 cells, there were less metastatic foci in the lungs compared with control animals (P = 0.0046; Fig. 5C, bottom). These results are supported by the majority of publications implicating a role for mutp53 in metastasis (20–23, 25–37, 48–50); however, mutp53 is dispensable for metastatic initiation, which requires Mdm2.

Here, we demonstrate that Mdm2 is extremely important in breast cancer metastases to the lung. Specifically, we show that Mdm2 is important to promote cancer invasiveness via cell migration, angiogenesis, and intravasation (Fig. 6). Molecularly, Mdm2 stabilizes HIF1α, which increases VEGF and transcription factors associated with intracellular metastatic programming (such as SNAI1; ref. 11). This intracellular pathway increases angiogenesis, and the decrease of vasculature observed with decreased Mdm2 provides one possible explanation for the small tumor size in the MDA468 shMdm2 tumors as well as the delay in tumor growth of TMD231 shMdm2 tumors. Therefore, the abundance of CTCs, and lung metastatic foci is reliant on the presence of Mdm2. Several studies have reported a role for Mdm2 in migration or EMT in vitro. In contrast, our experiments were conducted in vivo, which is significant in light of recent publications delineating the importance of the microenvironment in tumor growth and invasion.

In addition to being essential in early metastasis, Mdm2 may also be important in later stages as well. Although extravasation and colonization was not directly assessed, at least one study has shown that tail vein injection (bypassing the initial stages of metastasis) of transiently overexpressing Mdm2 in murine cells results in increased metastatic potential. This suggests that Mdm2 may have a role in extravasation and/or foci formation. Not surprisingly, this study also demonstrated increased metastatic potential with tail vein–injected murine cells that were transiently knocked down for p53. The main difference between this study and our study is that we knocked down gain-of-function p53 and not wild-type p53. The only other in vivo study that demonstrated a role for Mdm2 in metastasis was also in a wild-type p53 background and credited the decreased metastasis on the activation of p53 genes. Furthermore, the knockdown of Mdm2 by RNAi resulted in decreased proliferation and tumor growth, which also contributed to the decrease in metastasis in these wild-type p53 cells.

Our results provide a significant advancement in our understanding of how Mdm2 is regulating the cellular programs to promote tumor metastasis independent of gain-of-function mutp53. If mutp53 was driving tumorigenesis, one would expect the MDA468 shMdm2 tumors (which have significantly more mutp53 than the controls) to be increased in size. However, these cells grew much slower in the implanted animals. Furthermore, when mutp53 was knocked down, the tumors actually had increased neovascularization than the controls. Importantly, we show that the animals implanted with the shp53 cells had higher numbers of CTCs, suggesting that mutp53 is actually a disadvantage in the early stages of metastasis. However, the abundance of shp53 CTC did not correlate with increased numbers of metastatic foci, signifying that mutp53 is important for extravasation or other steps of metastasis. Finally, knocking down mutp53 by shRNA in shMdm2 cells did not alter EMT-related responses, demonstrating that the effects of Mdm2 knockdown are independent of mutp53 levels (Supplementary Fig. S3). Thus, we show there are independent and distinct yet dynamic roles of mutp53 and Mdm2 during the different phases of metastasis.

A major roadblock in the ability to treat metastatic disease is delineating the basic signaling pathways involved in the genesis, survival, drug resistance, migration, and growth of metastatic cells. The effectiveness of new therapies will require a comprehensive approach to target multiple pathways, preferably those that are important before tumor cells reach circulation. On the basis of our work, future drugs should be designed to interrupt the p53-independent functions of Mdm2.

No potential conflicts of interest were disclosed.

Conception and design: P.M. Hauck, C.N. Batuello, L.D. Mayo

Development of methodology: P.M. Hauck, C.N. Batuello, G.E. Sandusky

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): P.M. Hauck, E.R. Wolf, D.J. Olivos III, C.P. McAtarsney, R.M. Cournoyer, G.E. Sandusky, L.D. Mayo

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): P.M. Hauck, E.R. Wolf, D.J. Olivos III, C.N. Batuello, K.C. McElyea, R.M. Cournoyer, G.E. Sandusky, L.D. Mayo

Writing, review, and/or revision of the manuscript: P.M. Hauck, E.R. Wolf, D.J. Olivos III, C.N. Batuello, R.M. Cournoyer, L.D. Mayo

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K.C. McElyea

Study supervision: L.D. Mayo

The authors thank In Vivo Therapeutics Core for implantation and collection of in vivo data.

This study was supported by NIH/NCI CA172256 and a diversity supplement to this RO1, and Riley Children's Foundation (L.D. Mayo). This work was also supported by NIH: In Vivo Therapeutics Core IUSCC P30 CA082709, T32HL007910 (D.J. Olivos III), and T35HL110854 (R.M. Cournoyer).

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.