Cancer Research Audrey Hepburn Protein Translation and Cancer
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
Cancer Research Clinical Cancer Research
Cancer Epidemiology Biomarkers & Prevention Molecular Cancer Therapeutics
Molecular Cancer Research Cancer Prevention Research
Cancer Prevention Journals Portal Cancer Reviews Online
Annual Meeting Education Book Meeting Abstracts Online
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Kluger, H. M.
Right arrow Articles by Price, J. E.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Kluger, H. M.
Right arrow Articles by Price, J. E.
[Cancer Research 65, 5578-5587, July 1, 2005]
© 2005 American Association for Cancer Research

Molecular Biology, Pathobiology and Genetics

Using a Xenograft Model of Human Breast Cancer Metastasis to Find Genes Associated with Clinically Aggressive Disease

Harriet M. Kluger1, Dina Chelouche Lev4, Yuval Kluger3, Mary M. McCarthy1, Galina Kiriakova4, Robert L. Camp2, David L. Rimm2 and Janet E. Price4

Departments of 1 Medicine and 2 Pathology, Yale University School of Medicine, New Haven, Connecticut; 3 Department of Cell Biology, New York University, New York, New York; and 4 Department of Cancer Biology, University of Texas M.D. Anderson Cancer Center, Houston, Texas

Requests for reprints: Janet E. Price, University of Texas M.D. Anderson Cancer Center, 1515 Holcombe Boulevard, Box 173, Houston, TX 77030. Phone: 713-792-8577; Fax: 713-792-8747; E-mail: jprice{at}mail.mdanderson.org.


   Abstract
Top
 Abstract
Introduction
 Materials and Methods
 Results
 Discussion
 References
 
Metastasis is the primary cause of death from breast cancer. A xenograft model was used to identify genes potentially involved with metastasis, comparing expression in the poorly metastatic GI101A human breast cancer cell line and a highly metastatic variant, GILM2. cDNA microarray analyses of these isogenic variants were done using 16K Operon 70-mer oligonucleotide microarray slides. Differentially expressed genes were identified by ANOVA, and differences of ≥2.5-fold were found for 106 genes. Changes in protein or RNA expression were confirmed for 10 of 12 genes. Three markers, heat shock protein 70 (HSP-70), chemokine (C-X-C motif) ligand 1 (CXCL-1), and secreted leukocyte protease inhibitor (SLPI), were studied further with breast cancer tissue microarrays using a novel method of automated quantitative analysis. This uses cytokeratin to define pixels as breast cancer (tumor mask) within the tissue array spot and then measures intensity of marker expression using a cyanine 5–conjugated antibody within the mask. Scores were correlated with clinicopathologic variables. High HSP-70 expression and high nuclear CXCL-1 expression in primary tumors were both associated with decreased survival (P = 0.05 and 0.027, respectively). Expression of each marker was strongly associated with lymph node involvement (P = 0.0002, 0.008, 0.0012, and 0.012 for HSP-70, nuclear CXCL-1, cytoplasmic CXCL-1, and SLPI, respectively). Identification of genes associated with metastasis in experimental models may have clinical implications for the management of breast cancer, because some of these are associated with lymph node metastasis and survival and might be useful as prognostic markers or molecular targets for novel therapies.


   Introduction
Top
 Abstract
 Introduction
Materials and Methods
 Results
 Discussion
 References
 
Despite improvements in surgery and the use of adjuvant therapy, many patients still die from breast cancer. Metastatic disease, principally to the lungs, liver, bone, and brain, is the most common cause of breast cancer death (1). Therefore, there is still a need to find new and more effective treatments, and this can only be accomplished through a better understanding of the genes that contribute to disease progression. Two basic principles have become apparent from studies of the pathophysiology of metastasis. First, tumors are heterogeneous; thus, different cells from the same tumor can differ in their potential to form metastasis. Second, the metastatic process is not random but is a well-orchestrated sequence of events that depend on the properties of the tumor cells and their interactions with the microenvironment at the metastatic site (2, 3). The metastatic phenotype of breast cancer results from the accumulation of multiple genetic alterations. Some individual genes, such as HER-2/neu, p53, nm23, and cyclin D, determined in clinical and experimental studies of breast cancer, have prognostic or predictive value for disease outcome (4). In recent years, advances in high-throughput technologies have revolutionized the analysis of genetic events in tumor progression (5). Gene expression profiling significantly increases the number of genes that can be investigated and the speed of data collection. In addition, tissue microarrays enable us to simultaneously evaluate expression of a single protein in hundreds of tumors under uniform conditions (6).

Reproducing the complex interactions that occur during metastasis is difficult, and much of the current knowledge is based on xenograft models using orthotopic injection of human tumor cells in immunodeficient mice (2). Several investigators have combined experimental metastasis models and expression array analyses to identify genes associated with metastasis (7–9). Unfortunately, there are few human breast cancer cell lines that are tumorigenic and metastatic (10). We recently described the selection of a more aggressive GILM2 variant of the GI101A human breast cancer cell line by selection of cells from rare metastases in mice bearing tumors of the original cell line. GILM2 cells produce rapidly growing tumors and metastases in 90% to 100% of nude mice (11). In the present study, the GI101A and GILM2 variants were used to identify genes associated with metastasis by comparative cDNA microarray analysis based on the rationale that in a comparison of isogenic variants the genetic differences might be relatively few. Further, differences in gene expression are more likely to be related to differences in malignant behavior than if the study used cell lines from different individuals. The study identified 106 genes differentially expressed by ≥2.5-fold, with 59 increased and 47 decreased in the metastasis-derived GILM2 cell line. The difference in expression of 10 of 12 genes was confirmed by independent measurements of RNA and/or protein. Identification of genes associated with, or functionally involved, in breast cancer metastasis may have clinical implications for the management of breast cancer. The small heat shock protein (HSP) {alpha}B-crystallin was elevated in the metastasis-derived cells, and we have reported recently that expression of {alpha}B-crystallin is associated with lymph node metastasis and shorter survival in breast cancer patients (12). Here, we perform further studies to investigate whether other genes that are differentially expressed in the metastatic breast cancer cells have clinical prognostic or predictive value using a newly developed, objective method of automated quantitative analysis (AQUA) of tissue microarrays.


   Materials and Methods
Top
 Abstract
 Introduction
 Materials and Methods
Results
 Discussion
 References
 
Cell lines. The GI101A cell line was provided by the Goodwin Institute for Cancer Research, Inc. (Plantation, FL; ref. 13). The cells were maintained in monolayer culture in DMEM/F-12 medium supplemented with 10% fetal bovine serum (FBS), L-glutamine, and 5% v/v insulin-selenium-transferrin supplement (Sigma Chemical Co., St. Louis, MO). The GILM2 lung metastasis-derived variant was established as described previously (11).

Anchorage-independent growth. Single-cell suspensions of GI101A and GILM2 were mixed with 0.3% agarose and plated over base layers of 0.6% agarose, with 5 x 103 cells per 35 mm diameter well. The 0.3% agarose was prepared as described previously (14) and supplemented with 10% or 1% FBS. Cultures were incubated for 21 days and the numbers of colonies with >50 µm in diameter were counted.

Tumorigenicity and metastasis assays. Tumor cells in exponential growth phase were harvested by brief exposure to 0.25% trypsin in 0.02% EDTA and suspended in PBS, and 2.5 x 106 cells (0.1 mL) were injected into the mammary fat pad of female athymic NCr-nude mice (10). Mice were purchased from the National Cancer Institute-Frederick Cancer Research Facility (Frederick, MA). Tumor sizes were measured weekly using calipers and tumors of 1.5 cm diameter were surgically removed as described previously (11). Mice were killed 4 weeks after tumor removal and examined for metastases, visible macroscopically, or detected in sections of lungs and lymph nodes. The care and use of animals was approved by the Institutional Animal Care and Use Committee of the University of Texas M.D. Anderson Cancer Center (Houston, TX).

RNA extraction. Total RNA was extracted from cells using the TriReagent (Sigma Chemical) following the manufacturer's recommended protocol. The RNA was then washed using a Qiagen RNeasy kit (Qiagen, Valencia, CA). The samples were electrophoresed in a 1% agarose gel containing 2.2 mol/L formaldehyde to verify the integrity of the RNA.

Probe preparation and labeling. Probe preparation and labeling were done as described previously (15). Briefly, total RNA (75 µg) was used for each of the cell lines for each array. cDNA probes were synthesized by reverse transcription using SuperScript reverse transcriptase (Life Technologies/Invitrogen, Carlsbad, CA) with oligo(dT) and random hexamers as primers, incorporating allyl amine-dUTP (Sigma Chemical) into synthesized cDNA. Coupling of cyanine 3 (Cy3) or cyanine 5 (Cy5) dyes to allyl amine-dUTP–modified cDNA was done using NHS-ester Cy3 or Cy5 dye (Pharmacia, New York, NY) by incubation at 25°C for 90 minutes in subdued light.

cDNA microarray slide hybridization. Operon 70-mer 16K oligonucleotide microarray slides were purchased from the Keck Microarray Facility at Yale University (New Haven, CT). The array contained 0.7 x 1010 to 1.7 x 1010 oligonucleotide probes per gene for 16,000 genes. A list of the genes on the array is available on request. Hybridization of fluorescently labeled probes to the glass slides was done with hybridization buffer [50% deionized formamide, 12.5% SSPE (150 mmol/L NaCl, 10 mmol/L sodium phosphate, 1 mmol/L EDTA), 0.625% SDS, 1.5x Denhardt's with blockers (0.5 µg/µL mouse Cot-1 DNA, 0.1 µg/µL poly(A) (15A), 0.2 µg/µL yeast tRNA)] at 42°C for 18 to 24 hours. After hybridization, the slides were washed first with 1x SSC/0.1% SDS at 25°C for 15 minutes, then with 0.2x SSC/0.1% SDS, and finally with 0.2x SSC.

Microarray slide scanning. The slides were scanned with an Axon Model laser scanner. Analysis of the fluorescent hybridization signal of microarray slides was done with Genepix software, and the data were analyzed using Microsoft Excel.

Microarray experimental design. GI101A and GILM2 samples were hybridized together on oligonucleotide arrays four times. On the first two slides, GI101A was coupled with Cy3 and GILM2 with Cy5. On the third and fourth slides, the dyes were reversed.

Microarray data preprocessing. As a preliminary step in searching for differentially expressed genes between the two cell lines, we excluded genes for which we did not have good hybridization in at least one of the four measurements. Genes that had complete sets of four ratios with expression levels of >100 pixels per spot were selected for further analysis. In each of the four experiments, gene expression levels of the samples were normalized by the median expression level of all genes in that sample. These normalized expression levels were then used to calculate gene expression ratios between the two samples.

Differentially expressed genes. To identify the differentially expressed genes, we applied the ANOVA (16) package provided by the Jackson Laboratory (http://www.jax.org/research/churchill/software/anova/index.html) as described previously (15). We selected genes with a 2.5-fold increase or decrease in expression relative to GI101A with 99% confidence intervals. The four ratios per gene from the replicate hybridizations were then combined to form a composite ratio.

ELISAs. Culture supernatants were collected as described previously (17), and concentrations of chemokine (C-X-C motif) ligand 1 (CXCL-1) and interleukin-8 (IL-8) in the conditioned medium were measured by ELISA following the procedure recommended by the manufacturer of the reagents (R&D Systems, Minneapolis, MN).

Immunoblotting. Immunoblotting of total cell lysates was done as described previously (11). For detection of proteins in culture supernatants, serum-free medium was collected from cultures of GI101A and GILM2 incubated for 24 hours and concentrated in spin filters ~15-fold. Antibody hybridizations followed the manufacturer's recommended procedures. Antibodies were detected with horseradish peroxidase (HRP)–conjugated secondary antibodies and the Amersham enhanced chemiluminescence system (Amersham, Arlington Heights, IL). Immunoreactive bands were quantified by densitometry using ImageQuant software (Molecular Dynamics, Sunnyvale, CA). The primary antibodies used were monoclonal anti-{alpha}B-crystallin (Stressgen Biotechnologies, Victoria, British Columbia, Canada); polyclonal anti–insulin-like growth factor binding protein 3 (IGFBP3; Upstate Biotechnology, Inc., Lake Placid, NY); polyclonal anti-HSP-70 and polyclonal anti-Id-1 (Santa Cruz Biotechnology, Santa Cruz, CA); monoclonal anti–tissue inhibitor of matrix metalloproteinase (MMP) 1 (TIMP1; Oncogene Research Products, Boston, MA); and polyclonal anti–phospholipase A2 (cPLA2; Cell Signaling Technology, Beverly, MA).

Immunohistochemistry. Sections of paraffin-embedded tumors were used to compare protein expression in vivo. Immunohistochemistry was done as described previously (18) using antibodies to {alpha}B-crystallin, basic fibroblast growth factor (bFGF; Sigma Chemical), and secreted leukocyte protease inhibitor (SLPI; R&D Systems) and HRP-conjugated secondary antibodies. Antibody binding was detected using diaminobenzidine, and sections were counterstained with hematoxylin.

Measurements of RNA expression of selected genes. Total RNA was isolated using TriReagent following the manufacturer's recommended protocol. Levels of bFGF were measured using RNase protection as described previously (19). For quantitative PCR (Q-PCR), total RNA was reversed transcribed with random primers from the High Capacity cDNA Archive kit (Applied Biosystems, Foster City, CA). cDNA was amplified using the ABI 7000 Sequence Detection System for the expression of IL-8, connective tissue growth factor (CTGF), IGFBP3, SLPI, MMP-7, and TIMP-1 using Predeveloped TaqMan Assay Reagents or Assays-on-Demand gene expression products (Applied Biosystems). Results were recorded as mean threshold cycle (Ct), and relative expression was determined using the comparative Ct method as described previously (17) using human placenta RNA (Promega, Madison, WI) as a calibrator.

Tissue microarray construction. Expression of HSP-70, CXCL-1, and SLPI was studied on large cohort tissue microarrays using a newly developed method of AQUA of expression. Tissue microarrays were constructed as described previously (20). The arrays contained 324 node-positive and 331 node-negative breast cancer cases constructed from paraffin-embedded, formalin-fixed blocks obtained from the Yale University Department of Pathology archives. This cohort has been validated and described in prior publications (12, 20). Specimens were resected between 1962 and 1980, with follow-up between 4 months and 53 years. Complete treatment history was not available for the entire cohort. Most patients were treated with local irradiation. None of the node-negative patients were given adjuvant systemic therapy. A minority of the node-positive patients (~15%) received chemotherapy, and ~27% received tamoxifen (after 1978). The mean age of patients at diagnosis was 58 years. Slides were reviewed by a pathologist to select areas of invasive tumor to be placed on the tissue microarray using a Tissue Microarrayer (Beecher Instruments, Silver Spring, MD). Sections (5 µm thick) with diameters of 6 µm were cut and placed on glass slides using an adhesive tape system (Instrumedics, Inc., Hackensack, NJ) and UV cross-linking.

Fluorescent immunohistochemical staining. Staining was done for automated analysis of breast cancer specimens using a newly developed method as described previously (21). Briefly, slides were deparaffinized by rinsing with xylene followed by two changes of 100% ethanol and two changes of 95% ethanol. The slides were boiled in a pressure cooker containing 6.5 mmol/L sodium citrate (pH 6.0) for antigen retrieval. Endogenous peroxidase activity was blocked in methanol containing 2.5% H2O2 for 30 minutes at room temperature. After washing with TBS, the slides were incubated at room temperature for 30 minutes in 0.3% bovine serum albumin/1x TBS to reduce nonspecific background staining. Slides were incubated at 4°C overnight in a humidity tray with the primary antibodies [goat polyclonal anti-human HSP-70 antibody (1:1,200; Santa Cruz Biotechnology) and mouse anti-human CXCL-1 antibody (1:40) and goat anti-human SLPI (1:1,250; both from R&D Systems)]. Primary anti-cytokeratin antibodies, AE1/AE3, were used to identify tumor cells. For HSP-70 and SLPI, we used mouse anti-human cytokeratin at 1:200 (DAKO Corp., Carpinteria, CA), and for CXCL-1, we used rabbit anti-human cytokeratin at 1:200 (DAKO). After overnight incubation, slides were rinsed thrice in 1x TBS/0.05% Tween 20. Primary antibody identification was achieved as follows: for CXCL-1, slides were incubated at room temperature for 1 hour with goat anti-mouse HRP (Envision; DAKO) to identify the target and goat anti-rabbit IgG conjugated to Alexa 488 (Molecular Probes, Inc., Eugene, OR) at a dilution of 1:200 to identify the mask. 4',6-Diamidino-2-phenylindole (DAPI) was added simultaneously at 1:100 to visualize the nuclei. Slides were incubated for 10 minutes with Cy5 directly conjugated to tyramide at a 1:50 dilution (Perkin-Elmer, Boston, MA) for target antibody identification. For HSP-70 and SLPI, biotin-labeled anti-goat IgG (1:200, Vector Laboratories, Burlingame, CA), anti-mouse IgG conjugated to Alexa 488 (1:200), and DAPI were added for 1 hour. Streptavidin conjugated to HRP (1:200, Perkin-Elmer) was applied, and primary antibodies were labeled by adding Cy5 directly conjugated to tyramide (1:50). The slides were then rinsed in TBS/0.05% Tween 20 and coverslips were mounted.

Automated quantitative analysis. The tissue microarray slides were analyzed using AQUA as described previously and validated (21–25). When compared with traditional immunohistochemistry, this method is as good or better at predicting outcome based on expression of prognostic markers. Briefly, areas of tumor within the histospot were distinguished from stroma by creating a tumor mask with Alexa 488–tagged cytokeratin. Within the tumor mask, the nuclear compartment was identified using DAPI. The cytoplasmic/membranous compartment was identified by subtracting the DAPI (nuclear) signal from the tumor mask. The target markers, HSP-70, CXCL-1, and SLPI, were visualized with Cy5, and expression levels were measured for the two defined subcellular compartments within the tumor mask.

Image acquisition. Multiple monochromatic, high-resolution (1,024 x 1,024 pixels, 0.5 µm) grayscale images were obtained for each histospot using the x10 objective of an Olympus AX-51 epifluorescence microscope (Olympus, Melville, NY) set up with an automated Prior microscope stage and video and digital image acquisition driven by custom program and macro-based interfaces with IPLabs (Scanalytics, Inc., Fairfax, VA) software.

Algorithmic image analysis. Images were analyzed using algorithms that have been described previously (21–25). Formalin-fixed tissues can exhibit autofluorescence, causing background peaks. Two images (one in-focus and one out-of-focus) were taken of the compartment specific tags and the target marker. A percentage of the out-of-focus image was subtracted from the in-focus image for each pixel, to eliminate background, as described previously in detail (21). Subsequently, a second algorithm was used to assign each pixel in the image to a specific subcellular compartment and calculate the intensity of the signal in each location. Pixels that cannot accurately be assigned to a compartment were discarded. The data were saved and expressed as the average signal intensity per compartment area. For HSP-70 and SLPI, the signal was only cytoplasmic/membranous and no signal was detected in the nuclear compartment. For CXCL-1, signal was detected in both the nucleus and the cytoplasm. The data were analyzed separately for these two compartments, as with some other biomarkers, we have found differences in the predictive value between nuclear and cytoplasmic levels (26, 27). The target signal was measured on a scale of 0 to 255, and intensity was expressed relative to compartment area.

Statistical analysis. The Statview (SAS Institute, Inc., Cary, NC) software was used for data analyses. Continuous AQUA scores of target expression were divided into quartiles and associations with clinicopathologic variables were completed using the {chi}2 test. The prognostic significance of the variables was assessed for predictive value using the Cox proportional hazards model with breast cancer–specific survival as an end point. Patients who died of other causes or patients who were lost to follow-up were censored. Survival curves were generated using the Kaplan-Meier method.


   Results
Top
 Abstract
 Introduction
 Materials and Methods
 Results
Discussion
 References
 
Growth and metastatic potential of GI101A and GILM2. Anchorage-independent growth is associated with malignant potential (14), and the metastasis-selected GILM2 cells showed higher colony-forming efficiency than GI101A cells, most notably in cultures with reduced serum (Fig. 1A). When injected into the mammary fat pads of nude mice, the GILM2 cells showed a markedly increased growth rate compared with GI101A cells (Fig. 1B). The incidence of metastasis from the mammary fat pad tumors to lungs was significantly higher in mice bearing GILM2 tumors (Fig. 1C). The metastatic burden in the lungs of mice with GILM2 tumors was greater as reported previously (11). Thus, in both in vitro and in vivo assays, the GILM2 cells show a substantially more aggressive phenotype.



View larger version (12K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 1. Anchorage-independent growth, tumorigenicity, and metastasis of GI101A and GILM2 cells. A, numbers of colonies formed in cultures of 0.3% agarose supplemented with 1% or 10% FBS. No colonies grew in the cultures of GI101A supplemented with 1% FBS. Columns, mean of triplicate cultures; bars, SD. B, growth rates of tumors resulting from injection of 2.5 x 106 cells into the mammary fat pad of nude mice. C, incidence of metastases in the lungs of mice with mammary fat pad tumors (GI101A, n = 17; GILM2, n = 24). The incidence of metastases was significantly higher in mice bearing GILM2 tumors. P < 0.0001 (Fisher's exact test).

 
cDNA microarray analysis of differentially expressed genes in the high versus low metastatic potential cell lines. cDNAs from GI101A and GILM2 cells were hybridized on oligonucleotide arrays four times. Few differences were seen between the four hybridizations, with a high level of concordance across the arrays. Use of the ANOVA package identified 2,100 genes that were differentially expressed in a statistically significant manner (>99% CI). The expression of 890 genes was increased and 1,217 genes showed decreased expression in the GILM2 cells, giving a 13% change rate. However, the expression of most of these genes was altered between 1.1- and 1.5-fold (90% showed a <2-fold change). Only 106 genes with a ≥2.5-fold change in expression (0.66% of total genes on the array) were identified. Table 1 shows the 59 genes with ≥2.5-fold increased expression in the metastasis-selected GILM2 cells, and Table 2 lists the 47 genes with decreased expression. The genes are grouped in the tables according to function in the context of possible roles in breast cancer progression. Nine categories potentially linked to tumor aggressiveness were selected, including mediators of cell growth, antiapoptosis, motility, and angiogenesis. The final group of genes in each table have unknown function or are expressed sequence tags. Genes potentially promoting motility (e.g., CXCL-1) showed increased expression in the GILM2 cells. In contrast, the expression of genes involved in adhesion and cohesion (desmocolin and claudin-6) were reduced in the cells with increased metastatic potential.


View this table:
[in this window]
[in a new window]

  		Table 1. Genes with increased expression in the metastasis-selected GILM2 cells

 

View this table:
[in this window]
[in a new window]

  		Table 2. Genes with decreased expression in the metastasis-selected GILM2 cells

 
Validation of differences in gene expression. Twelve genes were selected for validation of the differences in expression in the microarray analyses using measurements of either RNA and/or protein. The genes selected included those with reported functions in breast cancer progression and those for which antibodies and real-time PCR primers were commercially available. Differences in expression, consistent with findings in the microarray analysis, were found for 10 of the 12 genes (Table 3). Increases in {alpha}B-crystallin, IGFBP3, CXCL-1, and cPLA2 proteins were measured by immunoblotting or ELISA. Increased MMP-7, CTGF, SLP1, and IGFBP3 expression in GILM2 cells was confirmed by Q-PCR. Reduced expression of HSP-70 in GILM2 cells was confirmed by immunoblotting, bFGF by RNase protection assay (RPA), and TIMP-1 by immunoblotting and Q-PCR. The difference in Id-1 expression found in the microarrays was not confirmed. The difference in IL-8 expression, measured by ELISA and Q-PCR, was neither unchanged nor decreased, respectively, in the GILM2 samples in these assays.


View this table:
[in this window]
[in a new window]

  		Table 3. Validation of differential gene expression in GI101A and GILM2 cells

 


View larger version (130K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 2. Immunohistochemical detection of {alpha}B-crystallin (A), SLPI (B), and bFGF (C) in GI101A and GILM2 xenograft tumors. More intense staining for {alpha}B-crystallin and SLPI was seen in the GILM2 tumors compared with GI101A tumors, whereas staining for bFGF was reduced in the GILM2 tumors. Original magnification, x100.

 
Validation of differences in expression of three genes in vivo was tested by immunohistochemistry using sections of tumors grown in immunodeficient mice. The intensity of staining for {alpha}B-crystallin and SLPI was higher in tumors of GILM2 compared with GI101A tumors. The staining for bFGF showed lower intensity in GILM2 tumors compared with GI101A tumors (Fig. 2). Thus, the immunohistochemistry results were consistent with the gene expression differences identified in the expression arrays.

Immunohistochemical staining of breast cancer tissue microarrays. HSP-70 was selected for further analysis on human tumors, as it is a HSP that was shown previously to be associated with poor outcome (28). CXCL-1 was selected based on a recent report of the association between chemokines and their receptors and breast cancer aggression (29). SLPI was selected, as it was found in a prior cDNA microarray study of a breast cancer cell line model to be up-regulated with increased invasion and metastasis (15). Of the 655 breast cancer samples on the tissue microarrays (324 node positive and 331 node negative), 504 (77%) were interpretable for HSP-70 staining, 402 (61%) for CXCL-1 staining, and 434 (66%) for SLPI staining. Spots deemed uninterpretable had insufficient tumor cells in the spot, loss of tissue in the spot, or an abundance of necrotic tissue. The staining intensity for HSP-70 ranged from 62 to 215, CXCL-1 from 8 to 150, and SLPI from 45 to 246. Survival information was available for 90% of the spots, and associations with survival and other clinical variables are presented below. Examples of strong and weak CXCL-1 staining are shown in Fig. 3.



View larger version (43K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 3. Low (A) and high (B) cytoplasmic and nuclear CXCL-1 expression in sample breast cancer histospots using cytokeratin to define the tumor mask (green), DAPI to define the nuclear compartment (blue), and Cy5 for CXCL-1 identification (red).

 
Survival analysis. Univariate survival analysis was done using the continuous AQUA scores for cytoplasmic/membranous staining of HSP-70, CXCL-1, and SLPI and nuclear staining of CXCL-1. Significant associations with decreased survival were found for HSP-70 (P = 0.05) and nuclear CXCL-1 staining (P = 0.027), whereas the associations with cytoplasmic CXCL-1 (P = 0.08) and SLPI (P = 0.95) were not significant. Continuous AQUA scores were then divided into quartiles, reflecting the use of routine statistical divisions in the absence of an underlying justification for division of expression levels. The AQUA quartiles were correlated with breast cancer–specific survival of the patients using 20-year follow-up. Kaplan-Meier survival curves with univariate Ps are shown in Fig. 4. These curves show that for HSP-70 and nuclear CXCL-1 increased expression is correlated with worse outcome. For HSP-70, the curves suggest a splitting of the data to define the top three quartiles as "high" expression and the first quartile as "low." For nuclear CXCL-1, the curves show overlap of the second and third quartiles, to define three distinct categories of high, medium, and low nuclear CXCL-1 expression. For cytoplasmic CXCL-1, the curves are similar to HSP-70, showing overlap of the top three quartiles (high expressers), which is separate from the lowest quartile. For SLPI, the curves for all four quartiles overlap. These designations are used for the remainder of the analyses. SLPI was divided into "high" and "low" expressers by the median AQUA score.



View larger version (35K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 4. Kaplan-Meier survival curves with Ps for HSP-70, CXCL-1 (nuclear and cytoplasmic), and SLPI. Continuous AQUA scores were divided into quartiles and correlated with breast cancer–specific survival of patients using 20-year follow-up information.

 
Clinicopathologic correlations and multivariate analyses. Using the Cox proportional hazards model, we did multivariate analyses to assess the independent predictive value of nuclear CXCL-1 and HSP-70 staining. We used the binary designations of "high" and "low" for HSP-70 and "high," "medium," and "low" for CXCL-1 as defined above. The other prognostic variables used to perform these analyses included tumor size, nodal status, estrogen receptor (ER) staining, progesterone receptor (PR) staining, HER-2/neu staining, nuclear grade, and patient age at diagnosis. Neither HSP-70 nor CXCL-1 retained their significance as independent predictors of survival, and the only three independent predictors of survival were nodal status, nuclear grade, and tumor size.

Correlation between clinicopathologic variables and marker staining was further examined, and the results are shown in Table 4. Strong HSP-70 and nuclear CXCL-1 staining was associated with negative ER expression (P = 0.0009 and 0.03, respectively), and strong CXCL-1 was associated with negative PR expression. The most striking associations were found between all four markers and lymph node involvement (P = 0.0002 for HSP-70, P = 0.0008 for nuclear CXCL-1, P = 0.0012 for cytoplasmic CXCL-1, and P = 0.012 for cytoplasmic SLPI).


View this table:
[in this window]
[in a new window]

  		Table 4. {chi}2 Ps of the association between markers and clinicopathologic variables

 
Significant associations were found between all four markers studied, with the weakest associations found between SLPI and nuclear or cytoplasmic CXCL-1 as shown in Table 4. As expected, high nuclear CXCL-1 expression was associated with high cytoplasmic CXCL-1, and both were associated with high HSP-70 (P < 0.0001).


   Discussion
Top
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
References
 
The transformation and progression of breast cancer is the result of multiple genetic changes. The development of expression array technologies, offering methods to examine patterns of altered gene expression across thousands of genes, has provided important tools for cancer research. Moreover, tissue microarrays allow for simultaneous measurement of selected markers in hundreds of patient specimens. Differential gene expression analyses have been used to define the expression patterns associated with malignant progression, metastasis, and poor prognosis for breast cancer (30, 31). In this report, the gene expression patterns in isogenic variants of a human breast cancer cell line with different metastatic potential were compared, and 106 genes (0.66% of the total analyzed) were identified as differentially expressed >2.5-fold. The relatively small number is likely due to the common origin of the cell lines, decreasing the genetic variation that may be found in comparisons of cell lines from different individuals (32). However, the fact that the expression levels of 106 genes were altered ≥2.5-fold in cells with increased metastatic potential indicates that multiple genetic events contribute to metastasis. The choice of cutoff at a 2.5-fold change in expression was somewhat arbitrary, and there is no consensus as to what should be considered a significant change in expression, with different values used in different studies. It is possible that a smaller change in expression of transcription factors may affect gene expression in the GILM2 variant, but these have been excluded from the current analysis by use of a 2.5-fold change cutoff. Different microenvironmental factors can also affect gene expression levels, and measurements of RNA from cultured tumor cells may not necessarily reflect expression in the tumors in vivo (33). We chose in this study to compare differences in gene expression in vitro, avoiding the influence of, or contamination with, tumor stroma but included analyses of xenograft tissues of the breast cancer variants in some of the validation experiments.

The reproducibility and accuracy of results from cDNA arrays should be considered, given the relatively high rate of false positives caused by biological variation, cross-reactivity with homologous sequences, random noise, and possible measurement errors (34, 35). Validation of the expression changes have been most commonly done with RNA, in Northern blots or real-time PCR, and the number of genes validated varies in different studies [e.g., 1 (32) to 15 (7)]. In this study, 10 of 12 genes showed a difference in either RNA and/or protein expression that agreed with results from the expression arrays. However, one limitation of RNA measurements is that they cannot assess the possible effect of post-translational modification, or of RNA stability, on protein expression and function(s). Several studies have shown that mRNA levels do not necessarily correspond with protein expression levels (36, 37).

Some of the genes identified in the array were among the "usual suspects" of metastasis-associated genes. These included TIMP-1, an inhibitor of MMPs, with reduced expression in GILM2 cells, and MMP-7, with increased expression in the metastasis-selected cells. Both TIMP-1 and MMP-7 have been implicated in invasion and metastasis (38, 39). Unlike other MMPs found in breast cancers, MMP-7 is expressed by the malignant cells rather than by stromal elements (40).

Several other genes previously linked with cancer progression and metastasis were among those with >2.5-fold change in this study. CXCL-1 and CXCL-2, two chemokines, were more highly expressed in GILM2. CXCL-1 is an autocrine growth factor for melanoma and promotes chemotaxis of granulocytes and endothelial cells through binding to CXCR2. It also promotes the migration of breast cancer cells in vitro (41) and tumor growth, metastasis, leukocyte infiltration, and angiogenesis in mouse squamous cell carcinoma (42). Elevated expression of SLPI in GILM2 cells was confirmed in validation experiments. This serine protease inhibitor has various activities, including a role in wound healing, anti-inflammatory processes, and cell proliferation (43). Consistent with the observation reported in this study, SLPI has been shown to promote the tumorigenicity and metastasis of mouse lung cancer cells by a currently unknown mechanism (44). A previous study using cDNA microarray analysis of another breast cancer cell line model showed SLPI to be up-regulated in the more aggressive daughter cell line when compared with the parental cell line (15).

Expression array analyses can identify genes that have not previously been linked with cancer or metastasis and may provide novel information. Few reports have documented expression of the small HSP {alpha}B-crystallin in breast cancer (45). This protein is thought to share chaperone functions of HSPs and may prevent apoptosis by binding to and preventing activation of caspase-3 (46, 47). Thus, increased expression in the metastatic variant GILM2 might confer a survival advantage. We reported previously the expression of {alpha}B-crystallin using the same tissue arrays used in this report and found a significant correlation between higher protein expression and lymph node metastasis, high nuclear grade, and shorter survival (12). In the same cohort, we evaluated the expression of HSP-70, CXCL-1 (nuclear and cytoplasmic), and SLPI. These proteins were highly coexpressed, particularly CXCL-1 and HSP-70. We found strong associations between high nuclear CXCL-1 and high HSP-70 expression and poor survival. Expression of each protein was associated with lymph node involvement. Because these markers are novel, cutoff points for "high/low" or "positive" versus "negative" have not been established. Regular brown stain for immunohistochemistry is flawed due to differences in results based on the concentration of the antibody used and subjectivity of the pathologist-based analysis. Use of our newly developed method of AQUA of expression allows for objective, continuous measurement of expression.

HSP-70 showed lower expression in GILM2 cells compared with GI101A cells in the gene array results, yet the tissue microarray results showed a strong association between high HSP-70 expression and tumor aggression. Although cell line models are useful screening tools for clinically relevant markers, the GI101A model only represents a single aggressive variant, established in an experimental setting, whereas the tissue microarrays contain specimens from hundreds of patients. Moreover, other studies in the literature support an association between high HSP-70 expression and tumor aggression and proliferation (28, 48). How HSP-70 may contribute to breast cancer progression is not well understood. A study of breast cancer specimens showed that the HSP expression was associated with increased proliferation (49), and similar to {alpha}B-crystallin, HSP-70 is reported to prevent the activation of caspases, blocking apoptosis (50) and thus providing a survival advantage.

Reliable predictors of lymph node involvement could change clinical practice once validated in a prospective fashion. About 30% of patients with newly diagnosed invasive breast cancer have positive sentinel lymph nodes at the time of initial surgery (51). Thus, accurate "molecular staging" of patients based on marker expression in primary tumors could ultimately replace costly and occasionally morbid sentinel lymph node dissections in ~70% of patients. Accurate molecular staging using predictors of lymph node involvement and survival could affect selection of adjuvant therapies in patients with stage I to III breast cancer. We now have many choices of adjuvant chemotherapy and antihormonal therapy, using different schedules and drug combinations, with variable degrees of toxicity. Selection of poor prognosis patients based on marker expression in primary tumors could enable us to reserve the most toxic therapies for those with the worst prognosis.

In conclusion, we used two variants of a human breast cancer cell line to identify genes differentially expressed in the more tumorigenic and metastatic variant. We studied the protein expression for three of these differentially expressed genes using human breast cancer specimens with associated clinical data. Two of the three (CXCL-1 and HSP-70) were associated with decreased survival, whereas all three were associated with lymph node metastases. Combined with the previous report using this model (12), four genes differentially expressed in the GI101A model of breast cancer progression were significantly associated with lymph node metastasis or shorter survival in breast cancer patients. Thus, cell line models can serve as a basis for further studies of the biology of breast cancer metastasis and provide models for further experimental analyses. The differentially expressed genes identified in this study might have an important role not only as prognostic markers but also as the next generation of molecular therapeutic targets.


   Acknowledgments
 
Grant support: Ethel F. Donaghue Women's Health Investigator Program at Yale, the Breast Cancer Alliance and the Susan G. Komen Breast Cancer Foundation (H.M. Kluger); the Yale Keck cDNA Microarray Facility is supported by the Anna and Argall Hull Fund, the C.G. Swebilius Trust, NIH grant 5 U24DK58776 (Dr. Kenneth Williams, Biotechnical Services, Yale University) and the Department of Pathology, Yale University; The Patrick and Catherine Weldon Donaghue Foundation for Medical Research and National Cancer Institute grant CA100825 (D.L. Rimm); NIH grant K0-8 ES11571 and the Breast Cancer Alliance (R.L. Camp); DAMD17-01-0455 from U.S. Army Medical Research and Materiel Command (J.E. Price); and National Cancer Institute Cancer Center support grant CA16672 to the University of Texas M.D. Anderson Cancer Center.

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.


   Footnotes
 
Note: H.M. Kluger and D. Chelouche Lev contributed equally to this work.

Received 1/13/05. Revised 4/14/05. Accepted 4/21/05.


   References
Top
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
 References
 

  1. Heimann R, Hellman S. Clinical progression of breast cancer malignant behavior: what to expect and when to expect it. J Clin Oncol 2000;18:591–9.[Abstract/Free Full Text]
  2. Fidler IJ. Critical determinants of metastasis. Semin Cancer Biol 2002;12:89–96.[CrossRef][Medline]
  3. Liotta LA, Kohn EA. The microenvironment of the tumour-host interface. Nature 2001;411:375–9.[CrossRef][Medline]
  4. Liu MC, Dickson RB, Lippman ME, Gelmann EP. Molecular oncology of breast cancer. In: Bland KI, Copeland EM III, editors. The breast: comprehensive management of benign and malignant disorders. 3rd ed. Philadelphia: Saunders; 2004. p. 499–535.
  5. Lockhart DJ, Winzeler EA. Genomics, gene expression and DNA arrays. Nature 2000;405:827–36.[CrossRef][Medline]
  6. Rimm DL, Camp RL, Charette LA, Costa J, Olsen DA, Reiss M. Tissue microarray: a new technology for amplification of tissue resources. Cancer J 2001;7:24–31.[Medline]
  7. Clark EA, Golub TR, Lander ES, Hynes RO. Genomic analysis of metastasis reveals an essential role for RhoC. Nature 2000;406:532–5.[CrossRef][Medline]
  8. Euer N, Schwirzke M, Evtimova V, et al. Identification of genes associated with metastasis of mammary carcinoma in metastatic versus non-metastatic cell lines. Anticancer Res 2002;22:733–40.[Medline]
  9. Kang Y, Siegel PM, Shu W, et al. A multigenic program mediating breast cancer metastasis to bone. Cancer Cell 2003;3:537–49.[CrossRef][Medline]
  10. Price JE. Metastasis from human breast cancer cell lines. Breast Cancer Res Treat 1996;39:93–102.[CrossRef][Medline]
  11. Chelouche Lev D, Kiriakova G, Price JE. Selection of more aggressive variants of the GI101A human breast cancer cell line: a model for analyzing the metastatic phenotype of breast cancer. Clin Exp Metastasis 2003;20:515–23.[Medline]
  12. Chelouche Lev D, Kluger HM, Berger AJ, Rimm DL, Price JE. {alpha}B-crystallin as a marker of lymph node involvement in breast cancer. Cancer 2004;100:2543–8.[CrossRef][Medline]
  13. Hurst J, Maniar N, Tombarkiewicz J, et al. A novel model of a metastatic human breast tumour xenograft line. Br J Cancer 1993;68:274–6.[Medline]
  14. Price JE. Clonogenicity and experimental metastatic potential of spontaneous mouse mammary neoplasms. J Natl Cancer Inst 1986;77:529–35.
  15. Kluger HM, Kluger Y, Gilmore-Hebert M, et al. cDNA microarray analysis of invasive and tumorigenic phenotypes in a breast cancer model. Lab Invest 2004;84:320–31.[CrossRef][Medline]
  16. Kerr MK, Martin M, Churchill GA. Analysis of variance for gene expression microarray data. J Comput Biol 2000;7:819–37.[CrossRef][Medline]
  17. Chelouche Lev D, Miller CP, Tellez C, Ruiz M, Bar-Eli M, Price JE. Different signalling pathways regulate VEGF and IL-8 expression in breast cancer: implications for therapy. Eur J Cancer 2004;40:2509–18.
  18. Yoneda J, Kuniyasu H, Crispens MA, Price JE, Bucana CD, Fidler IJ. Expression of angiogenesis-related genes and progression of human ovarian carcinomas in nude mice. J Natl Cancer Inst 1998;90:447–54.[Abstract/Free Full Text]
  19. Kim LS, Huang S, Lu W, Chelouche Lev D, Price JE. Vascular endothelial growth factor expression promotes the growth of breast cancer brain metastases in nude mice. Clin Exp Metastasis 2004;21:107–18.[CrossRef][Medline]
  20. Kluger HM, Dolled-Filhart M, Rodov S, Kacinski B, Camp RL, Rimm DL. Macrophage colony-stimulating factor-1 receptor expression is associated with poor outcome in breast cancer by large cohort tissue microarray analysis. Clin Cancer Res 2004;10:173–7.[CrossRef][Medline]
  21. Camp RL, Chung GG, Rimm DL. Automated subcellular localization and quantification of protein expression in tissue microarrays. Nat Med 2002;8:1323–7.[CrossRef][Medline]
  22. Camp RL, Dolled-Filhart M, King BL, Rimm DL. Quantitative analysis of breast cancer tissue microarrays shows that both high and normal levels of HER2 expression are associated with poor outcome. Cancer Res 2003;63:1445–8.[Abstract/Free Full Text]
  23. DiVito KA, Berger AJ, Camp RL, Dolled-Filhart M, Rimm DL, Kluger HM. Automated quantitative analysis of tissue microarrays reveals an association between high Bcl-2 expression and improved outcome in melanoma. Cancer Res 2004;64:8773–7.[Abstract/Free Full Text]
  24. Berger AJ, Camp RL, DiVito KA, Kluger HM, Halaban R, Rimm DL. Automated quantitative analysis (AQUA) of HDM2 expression in malignant melanoma shows association with early stage disease and improved outcome. Cancer Res 2004;64:8767–72.[Abstract/Free Full Text]
  25. Rubin MA, Zerkowski MP, Camp RL, et al. Quantitative determination of expression of the prostate cancer protein {alpha}-methyacyl-CoA racemase using automated quantitative analysis (AQUA): a novel paradigm for automated continuous biomarker measurements. Am J Pathol 2004;164:831–40.[Abstract/Free Full Text]
  26. Berger AJ, Kluger HM, Li N, et al. Subcellular localization of activating transcription factor 2 in melanoma specimens predicts patient survival. Cancer Res 2003;63:8103–7.[Abstract/Free Full Text]
  27. Dolled-Filhart M, Camp RL, Kowalski DP, Smith BL, Rimm DL. Tissue microarray analysis of signal transducers and activators of transcription 3 (Stat3) and phospho-Stat3 (Tyr705) in node-negative breast cancer shows nuclear localization is associated with a better prognosis. Clin Cancer Res 2003;9:594–600.[Abstract/Free Full Text]
  28. Thanner F, Sutterlin MW, Kapp M, et al. Heat-shock protein 70 as a prognostic marker in node-negative breast cancer. Anticancer Res 2003;23:1057–62.[Medline]
  29. Allinen M, Beroukhim R, Cai L, et al. Molecular characterization of the tumor microenvironment in breast cancer. Cancer Cell 2004;6:17–32.[CrossRef][Medline]
  30. Chang JC, Wooten EC, Tsimelzon A, et al. Gene expression profiling for the prediction of therapeutic response to docetaxel in patients with breast cancer. Lancet 2003;362:362–9.[CrossRef][Medline]
  31. van de Vijver MJ, He YD, van't Veer LJ, et al. A gene-expression signature as a predictor of survival in breast cancer. N Engl J Med 2002;347:1999–2009.[Abstract/Free Full Text]
  32. Liu D, Rudland PS, Sibson DR, Barraclough R. Identification of mRNAs differentially-expressed between benign and malignant breast tumour cells. Br J Cancer 2002;87:423–31.[CrossRef][Medline]
  33. Montel V, Pestonjamasp K, Mose E, Tarin D. Tumor-host interactions contribute to the elevated expression level of {alpha}1-antichymotrypsin in metastatic breast tumor xenografts. Differentiation 2005;73:88–98.[CrossRef][Medline]
  34. Hess KR, Zhang W, Baggerly KA, Stivers DN, Coombes KR. Microarrays: handling the deluge of data and extracting reliable information. Trends Biotechnol 2001;19:463–8.[CrossRef][Medline]
  35. Butte A. The use and analysis of microarray data. Nat Rev Drug Discov 2002;1:951–60.[CrossRef][Medline]
  36. Griffin TJ, Gygi SP, Ideker T, et al. Complementary profiling of gene expression profiling of gene expression at the transcriptome and proteome levels in Saccharomyces cerevisiae. Mol Cell Proteomics 2002;1:323–33.[Abstract/Free Full Text]
  37. Gygi SP, Rist B, Aebersold R. Measuring gene expression by quantitative proteome analysis. Curr Opin Biotechnol 2000;11:396–401.[CrossRef][Medline]
  38. Pacheco MM, Mourao M, Mantovani EB, Nishimoto IN, Brentani MM. Expression of gelatinase A and B, stromelysin-3 and matrilysin genes in breast carcinomas: clinico-pathological correlations. Clin Exp Metastasis 1998;16:577–85.[CrossRef][Medline]
  39. Brew K, Dinakarpandian D, Nagase H. Tissue inhibitors of metalloproteinases: evolution, structure and function. Biochim Biophys Acta 2000;1477:267–83.[CrossRef][Medline]
  40. Heppner KJ, Matrisian LM, Jensen RA, Rodgers WH. Expression of most matrix metalloproteinase family members in breast cancer represents a tumor-induced host response. Am J Pathol 1996;149:273–82.[Abstract]
  41. Youngs SJ, Ali SA, Taub DD, Rees RC. Chemokines induce migrational responses in human breast carcinoma cell lines. Int J Cancer 1997;71:257–66.[CrossRef][Medline]
  42. Loukinova E, Dong G, Enamorado-Ayalya I, et al. Growth regulated oncogene-{alpha} expression by murine squamous cell carcinoma promotes tumor growth, metastasis, leukocyte infiltration and angiogenesis by a host CXC receptor-2 dependent mechanism. Oncogene 2000;19:3477–86.[CrossRef][Medline]
  43. Ashcroft GS, Lei K, Jin W, et al. Secretory leukocyte protease inhibitor mediates non-redundant functions necessary for normal wound healing. Nat Med 2000;6:1147–53.[CrossRef][Medline]
  44. Devoogdt N, Hassanzadeh Ghassabeh G, Zhang J, Brys L, De Baetselier P, Revets H. Secretory leukocyte protease inhibitor promotes the tumorigenic and metastatic potential of cancer cells. Proc Natl Acad Sci U S A 2003;100:5778–82.[Abstract/Free Full Text]
  45. Pinder SE, Balsitis M, Ellis IO, Landon M, Mayer RJ, Lowe J. The expression of {alpha}B-crystallin in epithelial tumours: a useful tumour marker? J Pathol 1994;174:209–15.[CrossRef][Medline]
  46. Muchowski PJ, Clark JI. ATP-enhanced molecular chaperone functions of the small heat shock protein human {alpha}B crystallin. Proc Natl Acad Sci U S A 1998;95:1004–9.[Abstract/Free Full Text]
  47. Kamradt MC, Chen F, Sam S, Cryns VL. The small heat shock protein {alpha}B-crystallin negatively regulates apoptosis during myogenic differentiation by inhibiting caspase-3 activation. J Biol Chem 2002;277:38731–6.[Abstract/Free Full Text]
  48. Somiari RI, Sullivan A, Russell S, et al. High-throughput proteomic analysis of human infiltrating ductal carcinoma of the breast. Proteomics 2003;3:1863–73.[CrossRef][Medline]
  49. Vargas-Roig LM, Fanelli MA, Lopez LA, et al. Heat shock proteins and cell proliferation in human breast cancer biopsy samples. Cancer Detect Prev 1997;21:441–51.[Medline]
  50. LeBlanc AC. Natural cellular inhibitors of caspases. Prog Neuropsychopharmacol Biol Psychiatry 2003;27:215–29.[CrossRef][Medline]
  51. Krag D, Weaver D, Ashikaga T, et al. The sentinel node in breast cancer-a multicenter validation study. N Engl J Med 1998;339:941–6.[Abstract/Free Full Text]



This article has been cited by other articles:


Home page
 Cancer Res.Home page
H. H. Versteeg, F. Schaffner, M. Kerver, L. G. Ellies, P. Andrade-Gordon, B. M. Mueller, and W. Ruf
Protease-Activated Receptor (PAR) 2, but not PAR1, Signaling Promotes the Development of Mammary Adenocarcinoma in Polyoma Middle T Mice
Cancer Res., September 1, 2008; 68(17): 7219 - 7227.
[Abstract] [Full Text] [PDF]

Home page
 CarcinogenesisHome page
B. E. Bachmeier, I. V. Mohrenz, V. Mirisola, E. Schleicher, F. Romeo, C. Hohneke, M. Jochum, A. G. Nerlich, and U. Pfeffer
Curcumin downregulates the inflammatory cytokines CXCL1 and -2 in breast cancer cells via NF{kappa}B
Carcinogenesis, April 1, 2008; 29(4): 779 - 789.
[Abstract] [Full Text] [PDF]

Home page
 J EndocrinolHome page
R. Moral, R. Wang, I. H Russo, C. A Lamartiniere, J. Pereira, and J. Russo
Effect of prenatal exposure to the endocrine disruptor bisphenol A on mammary gland morphology and gene expression signature
J. Endocrinol., January 1, 2008; 196(1): 101 - 112.
[Abstract] [Full Text] [PDF]

Home page
 J. Biol. Chem.Home page
D. Chandra, G. Choy, and D. G. Tang
Cytosolic Accumulation of HSP60 during Apoptosis with or without Apparent Mitochondrial Release: EVIDENCE THAT ITS PRO-APOPTOTIC OR PRO-SURVIVAL FUNCTIONS INVOLVE DIFFERENTIAL INTERACTIONS WITH CASPASE-3
J. Biol. Chem., October 26, 2007; 282(43): 31289 - 31301.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Kluger, H. M.
Right arrow Articles by Price, J. E.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Kluger, H. M.
Right arrow Articles by Price, J. E.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
Cancer Research Clinical Cancer Research
Cancer Epidemiology Biomarkers & Prevention Molecular Cancer Therapeutics
Molecular Cancer Research Cancer Prevention Research
Cancer Prevention Journals Portal Cancer Reviews Online
Annual Meeting Education Book Meeting Abstracts Online