Plasmid Preparation

Each of the three retroviral plasmids, pBabe-neo-SV40 (10891), pBabe-hygro-hTERT (1773), pBabe-puro-RASV12 (1768) and the pseudotyping plasmid, pCL-VSVG (1733) were obtained from ADDGENE (Cambridge, MA) with each plasmid’s respective ID number in parentheses.¹ Plasmids were prepared according to the following procedures: Following bacterial culture and subsequent purification, purity of each plasmid was assessed by determination of the A260/A280 ratio. Once purity was verified, plasmid sequence integrity was verified using pBabe sequencing primer forward (5′-CTTTATC-CAGCCCTCAC-3′).

Retroviral Production and HMEC Infection

Transient retrovirus was produced according to the following procedures: Φnx cells were plated on gelatin-coated 100 mm plates and grown to a confluency of 85%. For each retroviral infection, 15 μg of pBabe plasmid and 15 μg of VSV-G plasmid was transfected into Φnx cells using Virapack Transfection Kit (Stratagene; Cedar Creek, TX). At 3 h post-transfection, cells were washed with PBS and fresh media added. At 48 h post-transfection retroviral supernatant was filtered through a 0.45 μm filter and Polybrene added to a final concentration of 4 μg/mL. Retroviral supernatant (2 mL) was then applied to HMECs at 35% confluency. After 3 h, 3 mL of MEGM was added and infection was carried out for 48 h. Cells were washed with PBS at 48 h postinfection to remove all traces of retrovirus and fresh MEGM was added. Infected HMECs were then allowed to recover from infection for 2 d before beginning selection. For each infection, antibiotic selection was carried out according to Elenbaas et al.²

Reverse Transcriptase-Polymerase Chain Reaction (RT-PCR) and Western Blot

RT-PCR was used to confirm expression of SV40-ER and exogenous hTERT (exo-hTERT). Total RNA was harvested using an RNEASY kit (Qiagen; Valencia, CA). Purified RNA was then subjected to DNase1 treatment and RNA repurified. To prevent RNA degradation, immediately following DNase1 treatment and subsequent purification, 2 μg of RNA was reverse transcribed to cDNA using cDNA First Strand Synthesis kit (Invitrogen; Carlsbad, CA). PCR was performed using primers, specific to SV40 Large T (F, 5′-GCTTTGCAAA-GATGGATAAAG; R, 5′-ACTAAACACAGCATGACTC), exo-hTERT (F, 5′-GACACACATTCCACAGGTCG; R, 5′-GACTCGACACCGT-GTCACCTAC), and Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) (F, 5′-GAGAGACCCTCACTGCTG; R, 5′-GATGGTA-CATGACAAGGTGC) cDNAs.¹ Electrophoresis was carried out at 100 V. Bands were visualized under UV light and analyzed with Kodak Digital Science Software (Kodak; Rochester, NY). Western blotting of protein extracts was used to determine stable expression of hRASV12. Nuclear protein extracts were isolated using NE-PER Nuclear Extraction Kit (Thermo Scientific; Rockford, IL). Protein concentration was determined with the Bradford method of protein quantification using the BioRad Protein Assay (BioRad; Hercules, CA). Nuclear protein extract (20 μg) was separated by electrophoresis at 100 V, transferred to a nitrocellulose membrane, and blocked in 5% dry milk in Tris buffered saline solution with 1% Tween (TBST) overnight. Primary antibody incubation was carried out overnight at 4 °C using mouse monoclonal antibodies specific to hRas (sc-53958). Membranes were then probed with goat anti-mouse IgG-HRP (sc-2005). Protein bands were visualized using enhanced chemiluminescence (ECL) reagents (Thermo Scientific; Rockford, IL). All antibodies used in immunoblotting were obtained from Santa Cruz Biotechnology (Santa Cruz, CA) with catalog numbers listed in parentheses.

Telomeric Repeat Amplification Protocol (TRAP)

A TRAP assay using TRAPeze Kit (Millipore Corporation; Billerica, MA) was performed to confirm activation of telomerase activity as a result of stable expression of exo-hTERT as previously described.¹³ TRAP assay was performed according to procedures described in the TRAPeze manual. Total protein extracts were harvested using CHAPs lysis buffer. Protein concentration was determined as described above and 250 ng of protein extract was used for each reaction. Heat-inactivated samples were incubated at 85 °C for 10 min. Protein extract from MDA-MB-231 cells served as a positive control and one reaction lacking protein extracts served as a negative control. Samples were incubated at 30 °C for 30 min, followed by a 33 cycle PCR. PCR conditions were followed according to the kit manual. Reaction products (25 μL) were loaded onto a 12.5% nondenaturing PAGE and electrophoresed for 1.5 h at 300 V. Gels were stained with SYBR Green (Lonza; Basel, Switzerland) and visualized under UV radiation.

THMEC Xenografts

Female Balb/c nude mice, 4–6 weeks of age, were obtained from Charles River Laboratories (Wilmington, MA). THMECs at 40 d post-hRas introduction were injected subcutaneously in the left flank of each mouse. For each injection, 0.1 mL of cell culture medium containing 1 × 10⁶ cells was mixed with an equal volume of Matrigel and the entire 0.2 mL was injected. Mice were sacrificed at 6 weeks postinjection and average tumor volume (mm³) was calculated as follows: tumor volume (TV) on day 42 = (a × b²)/2, where a is the largest diameter (in mm) and b is the diameter (in mm) perpendicular to a.

Experimental Design

We chose three points in the neoplastic progression to assess proteomic differences. These three time points in the neoplastic progression were as follows: SHMEC, THMEC 40 d post-hRAS, and THMEC 80 d post-hRAS. SHMEC cells are HMECs stably expressing SV40ER and exo-hTERT and represent premalignant breast cells. THMECs are HMECs stably expressing all three genetic elements, SV40ER, exo-hTERT, and hRAS-V12. For each time point, 4 biological replicates were generated for a total of 12 samples. For THMECs, two independent samples were generated for each time point (40 and 80 days) for each of two separate THMEC clonal lines generated from the same SHMEC clonal line. A schematic of the experimental design is shown in Supplementary Figure 1. THMECs at 80 days were derived from the same cell populations as those at 40 days. Four biological replicates of SHMEC were generated from one SHMEC clonal line, which was the same SHMEC cell line used to generate both THMEC cell lines. Experiments were designed in this fashion to optimize a balance between a sufficient number of biological replicates and a minimalization of variation and sample number. Finally, each sample was dye-swapped for a final n = 8 for each time point.

RNA and Protein Extraction for Real-Time PCR and 2D-DIGE

RNA, protein, and DNA were extracted using Tri Reagent (Ambion, Austin, TX) from HMEC-SV40-hTERT, THMEC 40 d post-hRas transfection, THMEC 80 d post-hRas, MCF10A, MCF10AT, MCF10CAa.1α, and MCF10CAd.1 following the manufacturer’s protocol. For each sample, 4 biological replicates were generated as described above. Following the final ethanol wash, the RNA pellet was resuspended in 100% deionized formamide and stored at −80 °C.¹⁴ Protein pellets were stored in 100% ethanol at −20 °C.

Preparation of Protein Extracts for 2D-DIGE and CyDye Labeling

Preparation of protein extracts and 2D-DIGE experiments were carried out as previously described by Kim et al.¹⁵ Precipitated protein from each cell sample was resuspended in isoelectric focusing (IEF) buffer and protein concentrations were determined using the 2D Quant Kit (GE Healthcare, Piscataway, NJ). All three CyDyes were obtained from GE Healthcare (Piscataway, NJ). To eliminate dye-specific bias, all samples were dye-swapped by labeling with both Cy3 and Cy5. An internal standard (IS) was generated by pooling 50 μg of each of the 12 samples and labeling all of this mixture with 200 pmol of Cy2. Protein (50 μg) for each sample was labeled according to the manufacturer’s instructions with 200 pmol of either Cy3 or Cy5. The Cy3 and Cy5 sample for each gel can be viewed in Table 1.

2D Difference Gel Electrophoresis-First Dimension Separation

For each 2D-DIGE gel, 50 μg of a Cy3-labled sample, 50 μg of a Cy5-labled sample, and 50 μg of the IS were pooled and the final volume brought up to 110 μL with IEF buffer before diluting 1:1 with rehydration buffer (7 M urea/2 M thiourea/4% CHAPS/40 mM Tris-HCl, pH 8.8) containing 1% IPGphor ampholytes, pI range 4–7 (GE Healthcare; Piscataway, NJ) and 30 mM dithiothreitol (DTT). Twelve 11.0 cm immobilized pH gradient (IPG) DryStrips pH gradient 4–7 (GE Healthcare; Piscataway, NJ) were rehydrated overnight for 20 h at room temperature (RT) in individual troughs in the IPG strip rehydrating chamber (GE Healthcare; Piscataway, NJ). The choice was made to use a pH 4–7 strip instead of the broader pH 3–10 strip based on previous experience of the authors and preliminary testing prior to the start of data collection.¹⁵ Each DryStrip was placed gel-side down on top of the respective 220 μL sample and overlaid with mineral oil. The following day, DryStrips were laid gel-side down and focused on an Ettan IPGphor II (GE Healthcare; Piscataway, NJ) at RT overnight using the following protocol: 6 h at 500 V, 1 h at 1000 V, 2.5 h at 6000 V, 2 h at 6000 V, 20 h at 300 V. Wicks were changed every 30 min for the first 1.5 h. The next day, DryStrips were placed at −80 °C.

2D Difference Gel Electrophoresis-Second Dimension Separation

DryStrips were thawed at RT for 15 min; and then equilibrated in SDS-sample buffer (6 M urea, 75 mM Tris-HCL, pH 8.8, 20% glycerol, 2% SDS, 65 mM DTT) twice for 15 min each and then once for 15 min in SDS-sample buffer containing freshly added iodoacetamide (135 mM) and a trace of bromophenol blue. Each strip was then placed on top of a 12.5% Criterion precast gel (Bio-Rad; Hercules, CA). Second-dimension separation for all 12 gels was carried out in a Dodeca Cell (Bio-Rad; Hercules, CA) at 150 V for 1.5 h at RT until bromophenol blue reached the bottom of the gel. Water chilled to 4 °C was circulated through the coils in the cell to maintain a low temperature.

Gel Image Acquisition

Following second-dimension separation, each gel was scanned on a Typhoon Trio+ Variable Mode Imaging System (GE Healthcare; Piscataway, NJ), using the specific laser band-pass filters for each dye’s excitation and emission wavelengths. The excitation/emission wavelength combinations were (480 ± 35 nm)/(530 ± 30 nm) for Cy2, (540 ± 25 nm)/(590 ± 35 nm) for Cy3, and (620 ± 30 nm)/(680 ± 30 nm) for Cy5. Each gel was scanned individually and Photo Multiplier Tube (PMT) voltages were adjusted for maximum image quality with minimal signal saturation and clipping. Images were cropped and exported as 16-bit GEL files using ImageQuant TL (GE Healthcare; Piscataway, NJ) and imported into DeCyder Image Analysis (GE Healthcare; Piscataway, NJ). During the scanning process, gel 3 tore (HMEC-SV40-hTERT replicate B-Cy3; THMEC 80 d replicate A-Cy5) and was excluded from image analysis.

Image and Statistical Analysis

GEL files were analyzed using the Batch Processor, Difference In-gel Analysis (DIA), and Biological Variation Analysis (BVA) modules within DeCyder Image Analysis Software. Batch processing was accomplished by setting 1250 as the upper spot limit. DIA was performed within individual gels comparing the Cy3 or Cy5 spot pattern to the Cy2 spot pattern. Ratios of Cy3/Cy5 spot volumes on each gel in the 11 gel data set were normalized against the Cy2 spot pattern. Results were obtained as abundance ratios for each spot. The following ratios were used: SHMEC/THMEC 40 d and THMEC 40 d/THMEC 80 d. To determine statistical significance, a Student’s t test was calculated for each ratio with Decyder’s False Discovery Rate applied. Spots determined as significantly different (FDR corrected p < 0.05) were then processed for identification. Expression graphs were generated as follows: the abundance for each spot was normalized to the internal standard (Cy2) on its gel to generate a standard abundance ratio for each spot. Abundance ratios for each spot at each time point were then averaged to generate a mean for SHMEC, THMEC 40 d, and THMEC 80 d.

Protein Identification

Following scanning, each gel was poststained with Sypro Ruby (Invitrogen; Carlsbad, CA) for spot picking. Spot picking was carried out using the ProPic (Isogen Life Science; Netherlands) robotic spot picker. Gel plugs were rinsed three times with 1 mL of 50% aqueous acetonitrile in 10 mM ammonium bicarbonate, pH 8.0, to remove Sypro Ruby stain. After evaporation of solvent by SpeedVac, each gel plug was rehydrated in 25 μL of 10 mM ammonium bicarbonate, pH 8.0. Trypsin Gold (10 μL of 12.5 μg/μL) was added to each gel plug solution, and the mixture was agitated overnight for 16 h. The supernatant was removed and plugs were washed twice with 10 mM ammonium bicarbonate. These rinses were combined, evaporated to dryness on a SpeedVac, then reconstituted in 10 μL of 0.1% formic acid. Peptide sequences were determined using an ABI-Sciex 4000 Qtrap (Applied Biosystems: Carlsbad, CA). The list of peptide sequences was exported and searched for matches within the nonredundant NCBI database (9/27/07) using the MASCOT search engine at www.matrix-science.com. Mass accuracy was set at 100 ppm, missed trypsin cuts at 1, Carbamidomethyl (C) fixed modifications, and Oxidation (HW) and Oxidation (M) variable modifications. The “gi” accession numbers obtained from MASCOT were used to search the UniProt Knowledgebase (UniProtKB) to obtain Swiss Protein Database ID numbers for each individual protein.

Analysis of Maspin mRNA and Protein Expression

Real-time PCR was used to assess mRNA levels of maspin (SERBINB5) in cell samples. RNA was prepared as described above using Trizol. RNA suspended in formamide was purified using RNEasy kit and resuspended in RNase free water. Purified RNA was reverse-transcribed as described above and cDNA was purified using a QiaQuick PCR Purification Kit (Qiagen; Valencia, CA). Purified cDNA (10 ng) was analyzed for expression of SERBINB5 (Hs00985283_m1) using Taqman Assays (Applied Biosystems; Carlsbad, CA) on a MiniOpticon Real-time Thermocycler (Bio-Rad; Hercules, CA). Specific Assay IDs are in parentheses. GAPDH served as the endogenous control. Expression values for THMEC 40 d and THMEC 80 d were calculated using the Ct method relative to HMEC-SV40-hTERT expression.¹⁶ Expression values for MCF10CAa.1α and MCF10CAd.1 were calculated using the ΔΔCt method relative to MCF10AT expression.¹⁶ Western blots were performed as described above to assess maspin protein levels in THMEC model and MCF10 model. Maspin (sc-22762) antibodies were obtained from Santa Cruz Biotechnology (Santa Cruz, CA), with catalog number in parentheses. Actin served as the internal control.

Cell Culture Observations and Assessment of Tumorigenicity

Figure 1D,E displays photographs taken of SHMECs and THMECs 30 d. Figure 1D shows SHMECs displaying a cobblestone-like appearance that is typical of an epithelial morphology. Figure 1E shows THMECs 30 d post-hRas. At 30 d post-hRas, THMECs have lost both contact inhibition and the cobblestone-like appearance seen in Figure 1D and now display a much more cylindrical phenotype indicated by the arrow in Figure 1E. This observation is supported by the fact that neoplastically transformed HMECs have been shown to undergo an epithelial to mesenchymal transition (EMT). We next assessed THMEC tumorigenicity in vivo by injecting THMECs at 40 d post-hRas introduction into immune-compromised nude mice. We choose 40 d post-hRas because at this point in the neoplastic induction foci formation was a common occurrence. All subcutaneous injections produced tumors (3 independent trials using 5 mice for a total n = 15). The mean tumor volume for all 15 mice was 126.22 mm³. A photograph of 3 representative tumors and the corresponding mice they were derived from can be viewed in Figure 1F. Arrows indicate the mice and location from which the tumors were removed. Injection of normal HMECs as a negative control was deemed unnecessary as HMECs are normal cells that lack the ability to form tumors.

Identification of Differentially Expressed Proteins during a Neoplastic Transformation

2D-DIGE was applied to assess differential expression at the proteomic level. A total of 4 protein samples for each of the following time points were analyzed: SHMEC, THMEC 40 d, and THMEC 80 d. We then dye-swapped each protein sample to remove any dye-specific bias, for a total n = 8 for each time point (4 bio reps, with 2 tech reps for each bio rep). All 24 labeled samples were separated two dimensionally on 12 gels, with each gel also containing an internal standard, generated by combining equal amounts of all proteins, for cross-gel comparison. A schematic of the images generated for each time point can be viewed in Figure 2A and the exact gel assignments can be viewed in Table 1. A Student’s t test (performed on both pairs; SHMEC vs THMEC 40 d and THMEC 40 d vs THMEC 80 d) and a One-Way ANOVA were used to identify spots that were significantly (FDR corrected p < 0.05) different in abundance over the course of neoplastic induction, in at least one of the three statistical tests performed. Selection of spots for identification was done by selecting the most significant spots for different trend types. Of the spots that were significantly different, 18 spots were chosen for identification. The location of each spot on a representative 2D gel can be viewed in Figure 2B. Spots identified to be differentially expressed are circled in pink. Of these 18 spots, 11 spots were identified as containing a single protein. These 11 spots are circled and numbered in Figure 2B and shown with their respective topographic peaks outlined in yellow. Additionally, 7 spots contained a mixture of proteins and further analysis is needed to identify which proteins are associated with the observed changes in expression. The 7 spots identified to have a mixture of proteins are circled but not numbered in Figure 2B. Spot numbers correspond to the master spot numbers assigned by Decyder. Topographic maps were exported from Decyder. Of these 11 spots, 2 spots had duplicate identities. In total, our methods produced 9 unique protein identifications (gene names are listed in parentheses): Heat Shock Protein 4A like (HSPA4L), Gelsolin (GSN), Keratin 1 (KRT1), T-complex Protein 1 (TCP1), FK506 binding protein 5 (FKBP5), Keratin 7 (KRT7), Maspin (SERPINB5), ribosomal protein P0 (RPLP0), and Stathmin (STMN1). Gelsolin and maspin each had two spots with individual IDs, most likely corresponding to differing post-translational modifications given the similarity in the migration in the second dimension. Proteomic data on all 18 spots can be viewed in Table 2, with spots containing a single protein in bold. Relative abundance graphs for the 11 spots containing a single protein are shown in Figure 3. FK506 binding protein 5, gelsolin, and stathmin increased in abundance during neoplastic induction, whereas heat shock protein 4-like, keratin 1, t-complex protein 1, keratin 7, maspin, and ribosomal protein P0 were decreased. For all spots that illustrated statistically significant change, with the exception of T-complex protein, the fold change (FC) > 1.20. Four spots displayed a near perfect linear down-regulation in protein abundance over 80 days as indicated by their respective coefficients of determination: keratin 7 (r² = 0.9913); maspin (r² = 0.9947); maspin (r² = 0.9997), ribosomal protein P0 (r² = 0.9971).

Open in a separate window

Figure 2

Two-dimensional difference gel electrophoresis of protein extracts from SHMEC, THMEC 40 d, THMEC 80 d. (A) Schematic of images per time point. Green and red gels correspond to Cy3 and Cy5 labeled samples, respectively. Blue arrow indicates the location of maspin (Master Spot #919) on a 2D-Gel image from each time point. (B) A representative image of a 2D gel. This image is of Gel 10 Internal Standard labeled with Cy2. Circled spots are those identified as being significantly (FDR corrected p < 0.05) differentially expressed between HMEC-SV40-hTERT/THMEC 40 d or HMEC-SV40-hTERT/THMEC 80 d, or ANOVA. Spot numbers for unique IDs correspond to the Master Spot numbers assigned by DeCyder. For each numbered spot, a topographic map of signal intensity is displayed with the area used to calculate intensity shown in yellow.

Open in a separate window

Figure 3

Protein abundance graphs of significantly differentially expressed proteins identified by 2D-DIGE. To the left of each graph is the corresponding gene name and master spot# assigned by Decyder. Protein expression results are displayed as mean log standard abundance ratios. See Materials and Methods for derivation of protein standard abundance ratios. Error bars reflect SEM Coefficients of Determination (r²) calculated using Microsoft Excel. (A) Heat Shock Protein 4A Like (HSP4AL); (B) Gelsolin (GSN); (C) Gelsolin (GSN); (D) Keratin 1 (KRT1); (E) T-Complex Protein 1A (TCP1); (F) FK506 binding protein 5 (FKBP5); (G) Keratin 7 (KRT7); (H) Maspin (SERPINB5); (I) Maspin (SERPINB5); (J) Ribosomal Protein P0 (RPLP0); (K) Stahmin (STMN1).

Analysis of SERPINB5 mRNA and Protein Expression in THMECs and Metastatic Progression Model

To validate our proteomic data, we used Western blotting to assess maspin (SERPINB5) protein expression and real-time PCR to assess SERPINB5 mRNA levels at all three time points of the THMEC model that were analyzed in 2D-DIGE (Figure 4A). Consistent with the proteomic data, SERPINB5 expression was progressively down-regulated over the course of tumorigenesis. The down-regulation also displayed the same linear trend as indicated by the coefficient of determination r² = 0.9849. SERPINB5 is known as a metastasis suppressor gene (MSG).¹⁸ MSGs suppress characteristics of metastasis such as invasion and motility, but have no effect on cellular division, immortality, or other nonmetastatic characteristics of cancer. The progressive down-regulation of SERPINB5 expression illustrates a trend toward a metastatic phenotype. To confirm that SERPINB5 is indeed associated with metastasis, we analyzed mRNA levels of SERPINB5 in three cell lines of the MCF10AT metastatic progression model (Figure 4B). MCF10AT is the premetastatic cell line, whereas MCF10CAa.1α. and MCF10CAd.1 are two metastatic clones each generated from separate mice by tail vein injections of MCF10AT.¹¹ The loss of SERPINB5 expression observed in THMECs can be observed in both metastatic clones relative to their premalignant origin (MCF10AT). SERPINB5 expression in both metastatic cell lines was roughly half of that observed in MCF10AT cells. Real-time data for SERPINB5 can be viewed in Figure 4A,B. We also assessed maspin protein levels in both models using Western blot analysis. Alterations in the protein expression of maspin in THMECs correlated with the observed changes in maspin mRNA expression. In the metastatic progression model, maspin expression was less in premalignant MCF10AT cells than in immortalized MCF10A cells. Maspin protein expression was undetectable in both metastatic carcinomas, MCF10CAd.1α and MCF10CAa.1. To summarize our findings on SERPINB5 expression, a near identical linear down-regulation over the entire 80 d period was observed in protein spots (#919 and #920), real-time assays, Western blotting, and microarray analysis (microarray data not shown) in THMEC model.

Open in a separate window

Figure 4

Real-Time PCR analysis of mRNA expression of maspin (SERBINB5) in THMECs and the MCF10AT Metastatic Progression Model. (A) mRNA expression of SERPINB5 in SHMEC, THMEC 40 d, 80 d. (B) mRNA expression of SERPINB5 in metastatic progression model (MCF10AT, MCF10CAd.1α, and MCF10CAa.). Expression ratios for THMEC 40 d, THMEC 80 d, MCF10CAd.1α, and MCF10CAa.1 were calculated using the ΔΔCt method relative to the expression of HMEC-SV40-hTERT (for THMECs) and MCF10AT (for MCF10CAd.1α and MCF10CAa.1). SHMEC and MCF10AT expression was set to 1.0. GAPDH which served as endogenous control. Error bars represent SEM. The columns represent mean relative expression. Coefficient of determination (r²) for maspin in THMEC model was calculated using Microsoft Excel. (C) Western blot of maspin protein expression in neoplastic progression model and metastatic progression model. Maspin protein expression was down-regulated over the course of tumorigenesis in THMECs. Maspin expression was less in premalignant MCF10AT cells than an in immortalized MCF10A cells. Maspin expression was undetectable in both metastatic carcinomas, MCF10CAd.1α and MCF10CAa.1. In both models, actin served as internal control.

The authors would like to acknowledge Lana Grinberg, Anna Pendleton, and Su Ngyuen for their help in generating THMECs. We would like to acknowledge the Targeted Metabolics and Proteomics Laboratory members Dr. Mark Cope, Gloria Robinson, and Richie Herring for assistance with the 2D-DIGE protocols. We would like to acknowledge the Tollefsbol lab members who provided informal advice during all aspects of the projects. J.T.D. was supported by an NCI Cancer Prevention and Control Training Program (R25CA047888). This work was also supported in part with grants from Susan G. Komen for the Cure (T.O.T., PI) and the NIH (R01 CA129415, T.O.T., PI). The instrumentation for running, imaging, and processing 2D-gels was provided by a NCRR Shared Instrumentation grant (S10 RR16849, H.K., PI). The mass spectrometers were purchased from funds provided by NCRR grants (S10 RR11329, RR13795, RR19231, S. Barnes, PI). Operation of the Comprehensive Cancer Center Proteomics-Mass Spectrometry Shared Facility was partially supported by funds from a P30 Core support grant (CA13148-35, E. Partridge, PI) to the UAB Comprehensive Cancer Center. We would like to acknowledge our collaboration with the lab of Dr. Danny Welch (University of Alabama at Birmingham) for providing us with the metastatic variants of the MCF10AT metastatic progression model.