Purpose: Inflammatory breast cancer (IBC) is associated with very poor prognosis. The aims of this study are (a) to prospectively identify differential gene expression patterns associated with IBC and (b) to confirm these pathways using tissue arrays.

Experimental Design: For gene expression analysis, IBC (n = 14) was clinically defined as rapid-onset cancer associated with erythema and skin changes, whereas non-IBC patients (n = 20) had stage III breast cancers, and cDNA analysis was carried out using the Affymetrix (Santa Clara, CA) HG-U133A microarrays. Tissue arrays were constructed from paraffin-embedded material, and the molecular phenotype of 75 IBC was compared with results from >2,000 non-IBC.

Results: Gene expression analyses indicated that IBC has higher expression of genes associated with increased metabolic rate, lipid signaling, and cell turnover relative to non-IBC tumors. Consistent with the expression analysis, IBC had statistically higher Ki-67 (93% versus 11%; P < 0.001). BAX expression, reflecting increased apoptosis and cell turnover, was significantly uniformly higher in almost all IBC (98% versus 66%; P < 0.05), whereas the expression of Bcl-2 was not significantly different. IBC tumors were more likely to be steroid hormone receptor negative (estrogen receptor, 49% versus 30%; P = 0.002; progesterone receptor, 68% versus 42%; P = 0.001). The expression of tyrosine kinases was not significantly different. E-cadherin was found to be expressed in 87% of IBC, whereas the expression p53 was not significantly different.

Conclusion: This study is one of the largest molecular analyses of IBC. Both IBC and non-IBC are genetically heterogeneous with consistent differences in the molecular phenotype of IBC.

Breast cancer is complex and heterogeneous, with a range of overlapping clinical phenotypes that manifest a wide variation in prognosis and outcome. Inflammatory breast cancer (IBC) is a rare form of invasive breast cancer characterized clinically by rapid onset of breast erythema, edema, and ridging, with the time from first symptom/sign to diagnosis of ≤3 months (1). Until the advent of systemic therapy, this disease was almost universally fatal. Approximately 20% of patients with IBC have gross distant metastases at the time of diagnosis (2), in contrast to <5% of invasive ductal carcinomas of no special type. The mean 5-year survival rate in most studies of patients with IBC with local therapy alone is <5%, with median survival from 12 to 36 months (3). This is in contrast to no-special-type breast cancer where the 5-year survival exceeds 50% with local treatment alone. Thus, with local therapy alone, the probability of surviving IBC is only one-tenth that of other invasive breast cancers. Hence, IBC provides an ideal clinical window to study the molecular events that contribute to metastases and poor survival outcome. However, little is known about the molecular biology of this unusual and lethal variant of invasive breast cancer. There has been limited research done on the genetic alterations involved in its pathogenesis and progression largely due to its rarity. Additionally, the incidence of IBC seems to be increasing, lending support to the idea that environmental factors may contribute to this phenotype (4).

The biology of IBC has been reported in small patient numbers, with lower rates of estrogen receptor (ER) and progesterone receptor (PgR) expression, and faster growth kinetics than no-special-type cancers (5). Other key genes involved in carcinogenesis, such p53, pS2, RhoC, and HER-2, have shown varied levels of alteration in IBC (6). Better understanding of the events that drive the unusually aggressive biology of this special type of breast cancer may help in discovering relevant and important genetic pathways of metastases. To obtain further insight into the tumor genetic mechanisms driving the rapid clinical progression of IBC and the underlying metastasis-promoting processes, we first identified IBC-associated pathways by analyzing differential breast tumor gene expression with cDNA microarrays. We then confirmed some of the molecular pathways associated with IBC by immunohistochemical analysis in a large cohort of patients identified from the Baylor Tumor Bank. The aim of this study is to begin to establish the molecular phenotype of IBC that may explain the extraordinary metastatic potential associated with IBC.

Patients

There are two different patient populations in this study. Group 1: Tumor samples were obtained from a total of 14 IBC patients and 20 non-IBC reference patients with stage III breast cancer, identified from two prospective neoadjuvant chemotherapy clinical trials carried out at the Baylor Breast Center from 1999 to 2004. IBC patients were clinically defined by two experienced breast medical oncologists (J.C. and R.E.) as erythema, skin edema (peau d'orange), ridging, and rapid onset from first symptom/sign to diagnosis of ≤3 months. Non-IBC reference patients had stage III breast cancers without erythema. Total mRNA was extracted from the 14 IBC and 20 non-IBC reference patients and hybridized to high-density cDNA arrays. Group 2: A total of 75 IBC were retrospectively identified from the Baylor Tumor Bank. Eligible patients were initially diagnosed between 1970 and 1991 with primary breast cancers and were treated with surgery and adjuvant radiation therapy. More than 250 hospitals submitted tumor specimens to this Bank originally for steroid hormone receptor analysis. To obtain demographic information (age and menopausal status) and extent of disease, individual hospitals were contacted and retrospective chart reviews were requested for each patient. Patients were diagnosed with IBC based on clinical criteria as designated by the referring institution, and at least 100 mg of frozen tumor powder were required to be available for molecular analysis. Based on these criteria, we analyzed the molecular phenotype of 75 IBC and compared their molecular phenotype with previously analyzed results of >2,000 non-IBC obtained from the same tumor bank.

Methods

Gene expression array experiments. Core biopsies were taken from a total of 14 IBC and 20 non-IBC reference tumors before administration of neoadjuvant chemotherapy, snap frozen, deidentified, and fractionated into RNA, DNA, and protein. Total RNA from the frozen core biopsy specimens was isolated according to the protocols recommended by Affymetrix (Santa Clara, CA) for GeneChip experiments. Total RNA extraction was done using Trizol reagent (Invitrogen Corp., Carlsbad, CA). To remove RNA fragments less than 200 nucleotides, which comprised 15% to 20% of total RNA, each sample was passed over a Qiagen RNeasy column (Qiagen, Valencia, CA). A chimeric oligo(dT)-T7 RNA polymerase promoter cDNA synthesis primer was then used to prepare double-stranded tumor cDNA. Reverse transcription was done on the cDNA to yield cRNA. Amplification and biotin labeling of antisense RNA was done using biotinylated ribonucleotides followed by cleanup of the reverse transcription. After chemical fragmentation, biotin-labeled cRNA was hydridized to Affymetrix HG-U133A microarrays. After automated washing and staining with streptavidin-phycoerythrin and biotinylated anti-streptavidin antibody (Vector Laboratories, Burlingame, CA), the arrays were scanned by Affymetrix GeneChip scanner (Agilent, Palo Alto, CA) and quantitated with Microarray suite (Affymetrix).

Statistical analysis. We used three software packages: dChip (http://www.dchip.org), BRB Array Tools (http://linus.nci.nih.gov/BRB-ArrayTools.html), and Array Analyzer (http://www.insightful.com). After scanning and low-level quantification using Microarray suite (http://www.affymetrix.com), we used dChip for normalization and estimation of expression values using the Li et al. PM-only model (7). The advantage of dChip is a model-based approach, which allows saving a model and applying it for individual new patient data. We used the PM-only model, because different studies found that it gives more precise expression estimations. We selected a subset of candidate genes by filtering on signal intensity to eliminate genes with uniformly low expression or genes whose expression did not vary significantly across the samples. BRB Array Tools and Array Analyzer were used for high-level analysis: identifying genes, Gene Ontology (GO) categories, and pathways differentially expressed between inflammatory and noninflammatory phenotypes.

We used t tests in BRB Array Tools (8) to find differentially expressed genes and two-way ANOVA in Array Analyzer (9) to account for the effect of ER and HER-2 status. Functional class scoring (10) was used to find GO categories and pathways that are significantly differentially expressed between inflammatory and noninflammatory phenotypes. This method is more powerful than common overrepresentation analysis of gene lists based on individually analyzed genes. It includes computing statistics that summarize Ps (for comparison of two classes) for all (N) genes in a GO category of pathways and comparison of these summary statistics with empirical distributions in random samples of N genes obtained by resampling. The two statistics used in pathway and GO comparisons were (a) LS statistics (defined as the mean negative natural algorithm of the Ps of the appropriate single genes univariate test) and (b) KS statistics (defined as the maximum difference between i/N and pi, where pi is the ith smallest P of the univariate test).

Immunohistochemical analysis. We compared differences in the molecular phenotype of the 75 IBC with previously analyzed results from >2,000 non-IBC reference tumors from the Baylor Tumor Bank. The assays for the IBC and non-IBC were done at different times. The pathways studied involved steroid hormone dependence (ER and PgR), apoptosis-related molecules (Bcl-2 and BAX), proliferation (Ki-67), growth factor receptor pathways [epidermal growth factor receptor (EGFR) and HER-2], p53, and cell adhesion molecules (E-cadherin). The standard method for immunohistochemical analysis was followed with slight modifications for tissue arrays. A small amount (∼100 mg) of fresh frozen pulverized tumor materials was formalin-fixed and then paraffin-embedded. From this material, 5 × 3–mm cores of cylinders of tissue were arranged 12 to a slide, and semiquantitative measurement of protein expression of multiple biomarkers was done. These tissue microarrays enabled semiquantitation of the expression of multiple proteins in a particular pathway or pathways to be studied simultaneously in a rapid and cost-efficient manner and facilitated the molecular profiling of many known proteins involved in cancer-related pathways. Table 1 summarizes the characteristics of the primary antibodies used. The slides were incubated with the primary antibodies as listed in Table 1, and secondary antibodies were then applied. The secondary antibodies were then linked to peroxidase-conjugated streptavidin. The chromogen signal was developed with 3,3′-diaminobenzidine, enhanced with osmium tetroxide, and contrasted to a 1% methyl green counterstain.

Table 1.

Antibodies for immunohistochemistry and cutoff for evaluation using the Allred et al. scoring system (11), where the total score = intensity + proportion score

Primary antibodyEvaluation cutoff points
ER Abbott (ER-IC) >2 
PgR Abbott (PR-ICA) >2 
HER-2 Zymed (tab 250) >5 
EGFR Triton (31G7) >0 
p53 Novocastra (DO7) >0 
Bcl-2 DAKO (124) >0 
BAX Zymed (2D2) >0 
E-cadherin Zymed (HECD-1) >0 
Ki-67 high Immunotech (Mib1) >9* 
Primary antibodyEvaluation cutoff points
ER Abbott (ER-IC) >2 
PgR Abbott (PR-ICA) >2 
HER-2 Zymed (tab 250) >5 
EGFR Triton (31G7) >0 
p53 Novocastra (DO7) >0 
Bcl-2 DAKO (124) >0 
BAX Zymed (2D2) >0 
E-cadherin Zymed (HECD-1) >0 
Ki-67 high Immunotech (Mib1) >9* 
*

% Positive cells.

Immunostained slides were evaluated under a light microscope by the study pathologists (S.M. and D.C.A.) without any knowledge of the patients' data. The signal was scored using a system estimating both the proportion and the average intensity of positive tumor cells as described previously (11). The proportion of positive staining cells on the entire slide was scored as follows: 0, none; 1, proportion <1:100; 2, proportion 1:100-1:10; 3, proportion 1:10-1:3; 4, proportion 1:3-2:3; 5, proportion >2:3. The intensity of the positive signal was scored as follows: 0, negative; 1, weak staining; 2, intermediate staining; 3, strong staining. The overall score was expressed as a summation of the proportion and intensity scores. Tumors were regarded as expressing the particular molecular marker if the overall score was >2 for ER and PgR, >0 for p53, >0 for Bcl-2, >0 for BAX, >0 for EGFR, >0 for E-cadherin, and >5 for HER-2. For Ki-67, the percentage of positive staining cells was determined by direct counting, and staining in >9% of tumor cells was considered high proliferation (12).

Gene expression array results. The characteristics of the 14 IBC patients and 20 non-IBC patients included in the gene expression array study are shown in Table 2. Of these, 13 IBC and 12 non-IBC samples were included in the final analysis after quality-control analysis. No statistically differences were found between IBC and non-IBC patients with respect to menopausal status, nodal status, grade, ER, PgR, or HER-2 status, although the statistical power to detect meaningful differences is relatively low due to small sample sizes. Considerable heterogeneity existed in both IBC and non-IBC, with weak signatures on individual gene-by-gene analyses and high false discovery rates. However, as patterns of gene expression rather than individual genes are more likely to discriminate between IBC and other cancers, we next did functional class scoring to discover GO categories and pathways that are significantly different between the two subtypes. For these analyses, statistical significance was defined as P ≤ 0.005.

Table 2.

Characteristics of the 25 patients included in the gene expression analysis

IBC (n = 13)Non-IBS (n = 12)P
Menopausal status    
    Premenopausal NS 
    Postmenopausal  
Node status    
    n = 0 NS 
    n > 0  
Grade    
    1-2 NS 
    3  
    ND  
ER status    
    Negative 11 NS 
    Positive  
PgR status    
    Negative 10 NS 
    Positive  
HER-2 status    
    Negative NS 
    Positive  
IBC (n = 13)Non-IBS (n = 12)P
Menopausal status    
    Premenopausal NS 
    Postmenopausal  
Node status    
    n = 0 NS 
    n > 0  
Grade    
    1-2 NS 
    3  
    ND  
ER status    
    Negative 11 NS 
    Positive  
PgR status    
    Negative 10 NS 
    Positive  
HER-2 status    
    Negative NS 
    Positive  

There were 16 pathways and 61 GO categories that significantly discriminated between IBC and non-IBC. The pathways involved in IBC were likely to have differential expression of fatty acid, glycerophospholipid, glycerosphingolipid, and inositol phosphate metabolism, bile acid and steroid biosynthesis, vesicle-associated cytoskeleton, extracellular matrix, glycolysis/glucogenesis, amino acid degradation and mTOR pathway, ribosome and mRNA processing, cell growth and differentiation, detoxification/cellular, stress and T-cell inflammatory response (Table 3). Consistent with the pathways, 61 significant GO categories discriminated between the two classes. These included genes involved primarily in metabolism and catabolism, cell turnover, amino acid metabolism, mRNA processing, protein transport, and immune response (Table 4). Although both sets of tumors overexpressed cell adhesion/signaling genes, IBC overexpressed the Alzheimer's disease–associated APP and γ-secretase constituents presenilin and anterior pharynx defective 1 homologue A, which also can proteolytically activate CDK5, Notch, Wnt, and HER4 pathway signaling. IBC also overexpressed extracellular matrix–associated CTSC, CTSF, CSPG2, and laminins relative to membrane cytoskeleton-associated integrins, EDIL3, SPTBN1, MAL, PROC, and KRT12 overexpressed in the non-IBC tumors. Overall, the results indicate higher metabolic rate, bioactive lipid signaling, and cell turnover of IBC relative to non-IBC tumors, with no statistically significant differences in steroid hormone receptor profiles.

Table 3.

Discriminatory Cancer Genome Anatomy Project pathways between IBC and non-IBC, with permutation P < 0.005

Pathway IDPathway descriptionNo. genesLS permutation PKS permutation P
KEGG: hsa00071 Fatty acid metabolism 73 1e−05 0.00935 
KEGG: hsa00280 Valine, leucine, and isoleucine degradation 45 1e−05 0.00448 
KEGG: hsa00650 Butanoate metabolism 33 9e−05 0.01313 
KEGG: hsa00062 Fatty acid biosynthesis (path 2) 16 0.00032 0.00126 
KEGG: hsa00053 Ascorbate and aldarate metabolism 10 0.00035 0.02172 
KEGG: hsa00640 Propanoate metabolism 39 0.00099 0.02776 
KEGG: hsa00120 Bile acid biosynthesis 27 0.00122 0.03894 
KEGG: hsa00310 Lysine degradation 40 0.00234 0.12269 
KEGG: hsa03020 RNA polymerase 30 0.00349 0.08216 
10 BioCarta: h_LairPathway Cells and molecules involved in local acute inflammatory response 15 0.00366 0.03092 
11 BioCarta: h_tcytotoxicPathway T cytotoxic cell surface molecules 22 0.00374 0.02253 
12 KEGG: hsa00330 Arginine and proline metabolism 52 0.00422 0.09767 
13 BioCarta: h_ephA4Pathway Eph kinases and ephrins support platelet aggregation 0.01216 0.00446 
14 KEGG: hsa00920 Sulfur metabolism 11 0.01323 0.00348 
15 BioCarta: h_PDZsPathway Synaptic proteins at the synaptic junction 23 0.0545 0.00039 
16 KEGG: hsa00910 Nitrogen metabolism 29 0.25434 0.004 
Pathway IDPathway descriptionNo. genesLS permutation PKS permutation P
KEGG: hsa00071 Fatty acid metabolism 73 1e−05 0.00935 
KEGG: hsa00280 Valine, leucine, and isoleucine degradation 45 1e−05 0.00448 
KEGG: hsa00650 Butanoate metabolism 33 9e−05 0.01313 
KEGG: hsa00062 Fatty acid biosynthesis (path 2) 16 0.00032 0.00126 
KEGG: hsa00053 Ascorbate and aldarate metabolism 10 0.00035 0.02172 
KEGG: hsa00640 Propanoate metabolism 39 0.00099 0.02776 
KEGG: hsa00120 Bile acid biosynthesis 27 0.00122 0.03894 
KEGG: hsa00310 Lysine degradation 40 0.00234 0.12269 
KEGG: hsa03020 RNA polymerase 30 0.00349 0.08216 
10 BioCarta: h_LairPathway Cells and molecules involved in local acute inflammatory response 15 0.00366 0.03092 
11 BioCarta: h_tcytotoxicPathway T cytotoxic cell surface molecules 22 0.00374 0.02253 
12 KEGG: hsa00330 Arginine and proline metabolism 52 0.00422 0.09767 
13 BioCarta: h_ephA4Pathway Eph kinases and ephrins support platelet aggregation 0.01216 0.00446 
14 KEGG: hsa00920 Sulfur metabolism 11 0.01323 0.00348 
15 BioCarta: h_PDZsPathway Synaptic proteins at the synaptic junction 23 0.0545 0.00039 
16 KEGG: hsa00910 Nitrogen metabolism 29 0.25434 0.004 

Abbreviation: KEGG, Kyoto Encyclopedia of Genes and Genomes.

Table 4.

GO categories that discriminate IBC and non-IBC, with permutation P < 0.005

GO categoryGO descriptionNo. genesLS permutation PKS permutation P
0003823 Antigen binding 34 1e−05 1e−05 
0006525 Arginine metabolism 0.00012 0.001 
0006787 Porphyrin catabolism 0.00019 1e−04 
0030333 Antigen processing 35 0.00036 1e−05 
0006527 Arginine catabolism 0.00036 0.00146 
0004129 Cytochrome c oxidase activity 26 0.00038 0.00304 
0015002 Heme-copper terminal oxidase activity 26 0.00038 0.00304 
0016675 Oxidoreductase activity, acting on heme group of donors 26 0.00038 0.00304 
0016676 Oxidoreductase activity, acting on heme group of donors, oxygen as acceptor 26 0.00038 0.00304 
10 0000051 Urea cycle intermediate metabolism 10 6e−04 0.00296 
11 0016813 Hydrolase activity, acting on carbon-nitrogen (but not peptide) bonds, in linear amidines 0.00086 0.00168 
12 0019882 Antigen presentation 44 0.00094 1e−05 
13 0050660 FAD binding 0.00103 0.00602 
14 0019883 Antigen presentation, endogenous antigen 17 0.00109 1e−05 
15 0019885 Antigen processing, endogenous antigen via MHC class I 17 0.00109 1e−05 
16 0000313 Organellar ribosome 25 0.00131 0.00315 
17 0005761 Mitochondrial ribosome 25 0.00131 0.00315 
18 0006778 Porphyrin metabolism 25 0.00154 0.01142 
19 0044270 Nitrogen compound catabolism 0.00159 0.01364 
20 0008147 Structural constituent of bone 0.00185 0.02226 
21 0046483 Heterocycle metabolism 57 0.00211 0.03156 
22 0050662 Coenzyme binding 23 0.00228 0.01673 
23 0006081 Aldehyde metabolism 12 0.00299 0.18169 
24 0004062 Aryl sulfotransferase activity 0.00335 0.00038 
25 0045012 MHC class II receptor activity 23 0.00361 1e−05 
26 0008194 UDP-glycosyltransferase activity 59 0.00393 0.03168 
27 0019835 Cytolysis 0.00394 0.07823 
28 0015020 Glucuronosyltransferase activity 11 0.00398 0.00913 
29 0005750 Respiratory chain complex III (sensu Eukaryota) 0.00398 0.00451 
30 0006743 Ubiquinone metabolism 0.00399 0.00398 
31 0006744 Ubiquinone biosynthesis 0.00399 0.00398 
32 0042375 Quinone cofactor metabolism 0.00399 0.00398 
33 0045426 Quinone cofactor biosynthesis 0.00399 0.00398 
34 0016782 Transferase activity, transferring sulfur-containing groups 32 0.00401 0.00066 
35 0005583 Fibrillar collagen 18 0.00422 0.00038 
36 0051017 Actin filament bundle formation 0.00429 0.14763 
37 0004029 Aldehyde dehydrogenase (NAD) activity 12 0.00429 0.17928 
38 0006354 RNA elongation 0.0043 0.03916 
39 0008146 Sulfotransferase activity 28 0.00452 0.00015 
40 0007585 Respiratory gaseous exchange 13 0.00474 0.0064 
41 0000050 Urea cycle 0.00481 0.02556 
42 0008021 Synaptic vesicle 20 0.00554 0.00043 
43 0008210 Estrogen metabolism 0.00574 0.00065 
44 0004075 Biotin carboxylase activity 0.00824 0.00037 
45 0007269 Neurotransmitter secretion 0.0111 0.00155 
46 0016421 CoA carboxylase activity 0.01226 0.00017 
47 0005540 Hyaluronic acid binding 17 0.01236 0.00244 
48 0016885 Ligase activity, forming carbon-carbon bonds 0.01367 0.00015 
49 0016706 Oxidoreductase activity, acting on paired donors, with incorporation or reduction of molecular oxygen, 2-oxoglutarate as one donor, and incorporation of one atom each of oxygen into both donors 0.01438 0.00125 
50 0006584 Catecholamine metabolism 14 0.01789 0.00275 
51 0018958 Phenol metabolism 14 0.01789 0.00275 
52 0005581 Collagen 34 0.01902 0.00085 
53 0000096 Sulfur amino acid metabolism 24 0.01908 0.00125 
54 0019884 Antigen presentation, exogenous antigen 17 0.03115 0.00044 
55 0004385 Guanylate kinase activity 13 0.04164 0.00396 
56 0016282 Eukaryotic 43S preinitiation complex 37 0.04445 0.00087 
57 0019886 Antigen processing, exogenous antigen via MHC class II 18 0.04481 0.00192 
58 0006941 Striated muscle contraction 18 0.05668 0.00382 
59 0006817 Phosphate transport 48 0.06297 1e−05 
60 0016805 Dipeptidase activity 0.07543 0.00398 
61 0015698 Inorganic anion transport 89 0.12696 0.00122 
GO categoryGO descriptionNo. genesLS permutation PKS permutation P
0003823 Antigen binding 34 1e−05 1e−05 
0006525 Arginine metabolism 0.00012 0.001 
0006787 Porphyrin catabolism 0.00019 1e−04 
0030333 Antigen processing 35 0.00036 1e−05 
0006527 Arginine catabolism 0.00036 0.00146 
0004129 Cytochrome c oxidase activity 26 0.00038 0.00304 
0015002 Heme-copper terminal oxidase activity 26 0.00038 0.00304 
0016675 Oxidoreductase activity, acting on heme group of donors 26 0.00038 0.00304 
0016676 Oxidoreductase activity, acting on heme group of donors, oxygen as acceptor 26 0.00038 0.00304 
10 0000051 Urea cycle intermediate metabolism 10 6e−04 0.00296 
11 0016813 Hydrolase activity, acting on carbon-nitrogen (but not peptide) bonds, in linear amidines 0.00086 0.00168 
12 0019882 Antigen presentation 44 0.00094 1e−05 
13 0050660 FAD binding 0.00103 0.00602 
14 0019883 Antigen presentation, endogenous antigen 17 0.00109 1e−05 
15 0019885 Antigen processing, endogenous antigen via MHC class I 17 0.00109 1e−05 
16 0000313 Organellar ribosome 25 0.00131 0.00315 
17 0005761 Mitochondrial ribosome 25 0.00131 0.00315 
18 0006778 Porphyrin metabolism 25 0.00154 0.01142 
19 0044270 Nitrogen compound catabolism 0.00159 0.01364 
20 0008147 Structural constituent of bone 0.00185 0.02226 
21 0046483 Heterocycle metabolism 57 0.00211 0.03156 
22 0050662 Coenzyme binding 23 0.00228 0.01673 
23 0006081 Aldehyde metabolism 12 0.00299 0.18169 
24 0004062 Aryl sulfotransferase activity 0.00335 0.00038 
25 0045012 MHC class II receptor activity 23 0.00361 1e−05 
26 0008194 UDP-glycosyltransferase activity 59 0.00393 0.03168 
27 0019835 Cytolysis 0.00394 0.07823 
28 0015020 Glucuronosyltransferase activity 11 0.00398 0.00913 
29 0005750 Respiratory chain complex III (sensu Eukaryota) 0.00398 0.00451 
30 0006743 Ubiquinone metabolism 0.00399 0.00398 
31 0006744 Ubiquinone biosynthesis 0.00399 0.00398 
32 0042375 Quinone cofactor metabolism 0.00399 0.00398 
33 0045426 Quinone cofactor biosynthesis 0.00399 0.00398 
34 0016782 Transferase activity, transferring sulfur-containing groups 32 0.00401 0.00066 
35 0005583 Fibrillar collagen 18 0.00422 0.00038 
36 0051017 Actin filament bundle formation 0.00429 0.14763 
37 0004029 Aldehyde dehydrogenase (NAD) activity 12 0.00429 0.17928 
38 0006354 RNA elongation 0.0043 0.03916 
39 0008146 Sulfotransferase activity 28 0.00452 0.00015 
40 0007585 Respiratory gaseous exchange 13 0.00474 0.0064 
41 0000050 Urea cycle 0.00481 0.02556 
42 0008021 Synaptic vesicle 20 0.00554 0.00043 
43 0008210 Estrogen metabolism 0.00574 0.00065 
44 0004075 Biotin carboxylase activity 0.00824 0.00037 
45 0007269 Neurotransmitter secretion 0.0111 0.00155 
46 0016421 CoA carboxylase activity 0.01226 0.00017 
47 0005540 Hyaluronic acid binding 17 0.01236 0.00244 
48 0016885 Ligase activity, forming carbon-carbon bonds 0.01367 0.00015 
49 0016706 Oxidoreductase activity, acting on paired donors, with incorporation or reduction of molecular oxygen, 2-oxoglutarate as one donor, and incorporation of one atom each of oxygen into both donors 0.01438 0.00125 
50 0006584 Catecholamine metabolism 14 0.01789 0.00275 
51 0018958 Phenol metabolism 14 0.01789 0.00275 
52 0005581 Collagen 34 0.01902 0.00085 
53 0000096 Sulfur amino acid metabolism 24 0.01908 0.00125 
54 0019884 Antigen presentation, exogenous antigen 17 0.03115 0.00044 
55 0004385 Guanylate kinase activity 13 0.04164 0.00396 
56 0016282 Eukaryotic 43S preinitiation complex 37 0.04445 0.00087 
57 0019886 Antigen processing, exogenous antigen via MHC class II 18 0.04481 0.00192 
58 0006941 Striated muscle contraction 18 0.05668 0.00382 
59 0006817 Phosphate transport 48 0.06297 1e−05 
60 0016805 Dipeptidase activity 0.07543 0.00398 
61 0015698 Inorganic anion transport 89 0.12696 0.00122 

We next investigated whether the 500 “intrinsic genes” first published by Sorlie et al. could discriminate different molecular subtypes of IBC and non-IBC (13). Unsupervised analysis revealed clusters in three major categories for both IBC and non-IBC: “ER” cluster, basal-like cluster, and HER-2 cluster (Fig. 1), reflecting that the heterogeneity and expression signatures inherent in non-IBC also exist with IBC.

Fig. 1.

Hierarchical clustering of expression data from Sorlie et al. Rows, genes; columns, samples. Expression level of each gene in a single sample is relative to its median level across all IBC (I) and non-IBC (N) samples. Red and blue, expression levels above and below the median, respectively. The dendrogram of samples represents overall similarities in gene expression profiles. Under the dendrogram, the horizontal colored boxes delimit the different subgroups: ER-like (red box), basal (green box), and HER-2-overexpressing (blue box). Branches of the core samples for each of the centroids are similarly color coded. Colored bars, right, locations of three gene clusters of interest: ER (red bar), basal (green bar), and HER-2 cluster (blue bar). Some genes included in these clusters are referenced by their HUGO abbreviation as used in Locus Link.

Fig. 1.

Hierarchical clustering of expression data from Sorlie et al. Rows, genes; columns, samples. Expression level of each gene in a single sample is relative to its median level across all IBC (I) and non-IBC (N) samples. Red and blue, expression levels above and below the median, respectively. The dendrogram of samples represents overall similarities in gene expression profiles. Under the dendrogram, the horizontal colored boxes delimit the different subgroups: ER-like (red box), basal (green box), and HER-2-overexpressing (blue box). Branches of the core samples for each of the centroids are similarly color coded. Colored bars, right, locations of three gene clusters of interest: ER (red bar), basal (green bar), and HER-2 cluster (blue bar). Some genes included in these clusters are referenced by their HUGO abbreviation as used in Locus Link.

Close modal

The dChip “genome” routine, which statistically assesses the map distribution of differentially expressed genes in IBC versus non-IBC, detected several significant “blocks” of nonrandom gene mapping at breast tumor loss of heterozygosity–associated loci, suggesting chromosomal or DNA copy number alterations (Fig. 2). Most, such as the chromosome 1q, 5q, and 6p “blocks,” are composed of genes from both tumor types, suggesting tumor genetic processes common to the IBC versus non-IBC tumors. However, two large “blocks” at breast cancer loss of heterozygosity–associated chromosome 14q24 and 19p13 loci are exclusively composed of genes overexpressed in IBC, suggesting that tumor genomic mechanisms contribute to the differential gene expression and phenotype of the these breast tumor types (Fig. 2).

Fig. 2.

dChip “genome” map showing the distribution of significant differentially expressed genes in IBC versus non-IBC. Significant “blocks” of nonrandom gene mapping at breast tumor loss of heterozygosity–associated loci were found at chromosome 14q24 and 19p13 loci (yellow boxes) in IBC, suggesting chromosomal or DNA copy number alterations.

Fig. 2.

dChip “genome” map showing the distribution of significant differentially expressed genes in IBC versus non-IBC. Significant “blocks” of nonrandom gene mapping at breast tumor loss of heterozygosity–associated loci were found at chromosome 14q24 and 19p13 loci (yellow boxes) in IBC, suggesting chromosomal or DNA copy number alterations.

Close modal

Immunohistochemical results. The development of tissue array technology provides methodology for high-throughput concomitant analyses of multiple proteins on many archival tumor samples. Following the gene expression results, we compared by immunohistochemistry a total of 75 IBC with >2,000 non-IBC reference breast cancers from the tumor bank for the following to reflect pathways involved in cell turnover, catabolism, and metabolism as well as steroid hormone receptors and known signaling pathways: ER, PgR, Ki-67, HER-2, EGFR, p53, Bcl-2, BAX, and E-cadherin. Table 5 summarizes the results obtained from the immunohistochemical experiments. Consistent with the expression analyses, IBC had statistically higher proliferation as measured by Ki-67 (93% of IBC samples had high proliferation versus 11% of non-IBC samples; P < 0.001). BAX expression, reflecting increased apoptosis and cell turnover, was significantly more frequent in almost all IBC (98% versus 66%; P < 0.05), whereas the expression of Bcl-2 was not significantly different (63% versus 68%). IBC tumors were more likely to be ER negative (49% versus 30%; P = 0.002) and PgR negative (68% versus 42%; P = 0.001). The expression of signaling tyrosine kinases was not significantly different (EGFR positive, 23% versus 19%; HER-2 positive, 26% versus 17%) between IBC and non-IBC samples. E-cadherin was found to be expressed in 87% of IBC but was not evaluated in non-IBC. In contrast, p53 was expressed significantly more often in non-IBC samples (52%) compared with IBC samples (32%; P = 0.002; Table 5).

Table 5.

Tissue arrays comparing 75 IBC with >2,000 non-IBC: immunohistochemistry results

Inflammatory (n = 75), %Non-IBC (n = 2,093), %P
ER− 49 30 0.002 
PgR− 68 43 0.001 
HER-2+ 17 26 0.09 
EGFR+ 23 19 0.4 
p53 expression 32 52 0.002 
Bcl-2 expression 63 68 0.5 
BAX expression 98 66* <0.05 
E-cadherin expression 87   
High Ki-67 93 11 0.001 
Inflammatory (n = 75), %Non-IBC (n = 2,093), %P
ER− 49 30 0.002 
PgR− 68 43 0.001 
HER-2+ 17 26 0.09 
EGFR+ 23 19 0.4 
p53 expression 32 52 0.002 
Bcl-2 expression 63 68 0.5 
BAX expression 98 66* <0.05 
E-cadherin expression 87   
High Ki-67 93 11 0.001 
*

n = 200.

E-cadherin expression was not done on non-IBC.

This study is one of the largest comprehensive molecular analyses of IBC, using both gene expression arrays and tissue arrays. The results showed that both IBC and non-IBC tumors are genetically heterogeneous, with molecular phenotypes as previously described inherent in both tumor types (13), and some consistent differences exist in IBC compared with non-IBC.

Gene expression analyses indicated that genomic mechanisms contribute to the phenotype of IBC, with nonrandom regions of differential genes associated with IBC. IBC has also higher expression of genes associated with higher metabolic rate, bioactive lipid signaling, and cell turnover of IBC relative to non-IBC tumors. Consistent with these results, tissue array analyses confirmed that IBC tumors were rapidly proliferating, with nearly all tumors showing elevated Ki-67, and with increased apoptotic markers like BAX. Ki-67 is a nuclear endogen expressed only in proliferating cells (late G1, S, M, and G2) and is associated with poorer prognosis (14). Nearly all IBC had a high proliferation fraction, 93% versus 11% of non-IBC. This is also consistent with the clinical observation of rapid onset in IBC together with poorer prognosis and outcome compared with other locally advanced breast cancers. Consistent with increased cell turnover, BAX was significantly increased in IBC. The proapoptotic activity of BAX is neutralized when BAX is bound to Bcl-2 or other members of the Bcl-2 family (e.g., Bcl-XL; refs. 15, 16). From these results, this apoptotic pathway may be important therapeutic target in improving outcome in patients with IBC.

The p53 tumor suppressor gene has several functions affecting cell turnover, including activating the cell cycle by activating p23 and also promoting apoptosis (17). Mutation and loss of p53 function are associated with high proliferation rates and poor clinical outcome. Based on the high proliferation rates in IBC, we might have predicted p53 accumulation in IBC. Contrary to this expectation, we found p53 accumulation in approximately one-third of IBC cancers and half of non-IBC tumors (32% versus 52%). These results may indicate that the high cell turnover in IBC could be driven by mechanisms other than altered p53 or that p53 may be excluded from the nucleus and therefore may be nonfunctional (18).

Gene expression analyses did not show higher expression of inflammatory components in IBC compared with non-IBC group, consistent with the thinking that the erythema in IBC results from blockage of the lymphatic channels with tumor emboli rather than direct infiltration of the skin by inflammatory cells (19). Our results are consistent with other studies showing expression of E-cadherin in ∼90% of IBC (20). E-cadherin is thought to be a tumor suppressor gene, and its absence or low expression has been associated with high histologic grade, increased invasiveness, and high metastatic potential (21). Its conserved expression in IBC, although unexpected, has been described previously (20, 22, 23). A possible explanation for this paradoxical conservation of E-cadherin might be that the loss of E-cadherin is only transient and associated with the transit of cancer cells into the lymphovascular system. Once in the circulation, the cancer cells may reinstate the expression of E-cadherin, thereby facilitating intracellular adhesion and the formation of tumor emboli (19). Our observation that E-cadherin is frequently expressed in IBC might support the possibility that its expression may be important in the formation of tumor emboli.

Surprisingly, RhoC was not overexpressed in our group of IBC tumors, as published previously (6), but are, however, consistent with results by Bieche et al. (24). RhoC GTPase is a member of the Ras superfamily of small GTP-binding proteins. It has been shown to enhance invasiveness and motility, and this apparent discrepancy may arise from the genetic heterogeneity of IBC.

The importance of steroid hormone receptors has been well established in breast cancer. Estrogen, acting through the ER, promotes breast cancer growth and development. ER is expressed in 60% to 70% of invasive breast cancers and is associated with low proliferation rate, better response to endocrine therapy (25), and a lower risk of relapse (26, 27). Interestingly, immunohistochemical analysis showed that although IBC was slightly more likely to be ER negative and PgR negative, a third to a half still expressed these receptors. The growth factor receptors EGFR and HER-2 are often associated with poorer prognosis and increased tumor proliferation (28, 29), but from our results it is unlikely that the EFGR/HER-2 family is primarily responsible for the increased proliferation of IBC.

One of the limitations of this project is in the selection of IBC from the Baylor Tumor Bank. It is uncertain how many of the 75 IBC cases from the tumor bank actually fulfilled the clinical goal standard definition of IBC, because these cases represent assignment of IBC by the referring “community” physicians. Another limitation is that the non-IBC samples may not be true controls, because a portion of these may have included lower-grade and early-stage cancers.

We compared our results with two previously published differentially expressed genes for IBC (22). Of the 90 differentially expressed genes in the Bertucci et al. article (22), only kinase anchor protein 1 (PRKA) was significantly differentially expressed in our data set. Of the 50 differentially expressed genes from the Van Laere et al. article (30), only 5 genes overlapped (hemochromatosis, platelet-derived growth factor-α polypeptide, selectin P ligand, nucleotide binding protein 1, and cell division cycle 25B). Interestingly, little overlap in gene expression was observed between these two earlier studies, further supporting our hypothesis that IBC express a hyperproliferative profile with genetic heterogeneity.

Breast cancer is a clinically diverse disease, and this diversity is driven by multiple genetic alterations and molecular events. IBC, with its poor prognosis and distinct clinical manifestation, is a hallmark of this genetic diversity. Using small amounts of material, we have used tissue arrays and gene expression analysis to examine the tumor-genetic mechanisms that might contribute to the extraordinary malignancy of IBC. Mirroring their distinct clinical behavior, IBC have a profile that is hyperproliferative and seem to be driven by several diverse pathways. The Bcl-2/BAX apoptotic pathway may be an important target. From our results, pathways especially those involved in fatty acid and lipid metabolism seem to be important, and interventions, including life-style modifications, may improve outcome. We are continuing to refine this profile further and to investigate its implications for biology and treatment.

Grant support: Breast Cancer Research Foundation, Emma Jacobs Clinical Breast Cancer Fund, and National Cancer Institute Breast Cancer Specialized Programs of Research Excellence grant P50 CA50183.

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.

1
Lee B, Tannenbaum N. Inflammatory carcinoma of the breast: a report of twenty-eight cases from the breast clinic of memorial hospital.
Surg Gynecol Obstet
1924
;
39
:
580
–5.
2
Levine PH, Steinhorn SC, Ries LG, Aron JL. Inflammatory breast cancer: the experience of the Surveillance, Epidemiology, and End Results (SEER) program.
J Natl Cancer Inst
1985
;
74
:
291
–7.
3
Jaiyesimi IA, Buzdar AU, Hortobagyi G. Inflammatory breast cancer: a review.
J Clin Oncol
1992
;
10
:
1014
–24.
4
Hance KW, Anderson WF, Devesa SS, Young HA, Levine PH. Trends in inflammatory breast carcinoma incidence and survival: the surveillance, epidemiology, and end results program at the National Cancer Institute.
J Natl Cancer Inst
2005
;
97
:
966
–75.
5
Paradiso A, Tommasi S, Brandi M, et al. Cell kinetics and hormonal receptor status in inflammatory breast carcinoma. Comparison with locally advanced disease.
Cancer
1989
;
64
:
1922
–7.
6
van Golen KL, Wu ZF, Qiao XT, Bao L, Merajver SD. RhoC GTPase overexpression modulates induction of angiogenic factors in breast cells.
Neoplasia
2000
;
2
:
418
–25.
7
Li C, Wong WH. Model-based analysis of oligonucleotide arrays: expression index computation and outlier detection.
Proc Natl Acad Sci U S A
2001
;
98
:
31
–6.
8
Wright GW, Simon RM. A random variance model for detection of differential gene expression in small microarray experiments.
Bioinformatics
2003
;
19
:
2448
–55.
9
Benjamini Y, Drai D, Elmer G, Kafkafi N, Golani I. Controlling the false discovery rate in behavior genetics research.
Behav Brain Res
2001
;
125
:
279
–84.
10
Pavlidis P, Qin J, Arango V, Mann JJ, Sibille E. Using the Gene Ontology for microarray data mining: a comparison of methods and application to age effects in human prefrontal cortex.
Neurochem Res
2004
;
29
:
1213
–22.
11
Allred DC, Harvey JM, Berardo M, Clark GM. Prognostic and predictive factors in breast cancer by immunohistochemical analysis.
Mod Pathol
1998
;
11
:
155
–68.
12
Gerdes J, Li L, Schlueter C, et al. Immunobiochemical and molecular biologic characterization of the cell proliferation-associated nuclear antigen that is defined by monoclonal antibody Ki67.
Am J Pathol
1991
;
138
:
867
–73.
13
Sorlie T, Perou CM, Tibshirani R, et al. Gene expression patterns of breast carcinomas distinguish tumor subclasses with clinical implications.
Proc Natl Acad Sci U S A
2001
;
98
:
10869
–74.
14
Wintzer H, Zipfel I, Schulte-Mounting J, Hellerich U, von Kleist S. Ki67 immunostaining in human breast tumours and its relationship to prognosis.
Cancer
1991
;
67
:
421
–8.
15
Krajewski S, Mai JK, Krajewska M, Sikorska M, Mossakowski MJ, Reed JC. Upregulation of bax protein levels in neurons following cerebral ischemia.
J Neurosci
1995
;
15
:
6364
–76.
16
Berardo MD, Elledge RM, de Moor C, Clark GM, Osborne CK, Allred DC. bcl-2 and apoptosis in lymph node positive breast carcinoma.
Cancer
1998
;
82
:
1296
–302.
17
Polyak K. Is p53 a breast cancer gene?
Cancer Biol Ther
2002
;
1
:
37
–8.
18
Moll UM, Riou G, Levine AJ. Two distinct mechanisms alter p53 in breast cancer: mutation and nuclear exclusion.
Proc Natl Acad Sci U S A
1992
;
89
:
7262
–6.
19
Kleer CG, van Golen KL, Merajver SD. Molecular biology of breast cancer metastasis. Inflammatory breast cancer: clinical syndrome and molecular determinants.
Breast Cancer Res
2000
;
2
:
423
–9.
20
Charafe-Jauffret E, Tarpin C, Bardou VJ, et al. Immunophenotypic analysis of inflammatory breast cancers: identification of an “inflammatory signature.”
J Pathol
2004
;
202
:
265
–73.
21
Parker C, Rampaul RS, Pinder SE, et al. E-cadherin as a prognostic indicator in primary breast cancer.
Br J Cancer
2001
;
85
:
1958
–63.
22
Bertucci F, Finetti P, Rougemont J, et al. Gene expression profiling identifies molecular subtypes of inflammatory breast cancer.
Cancer Res
2005
;
65
:
2170
–8.
23
Bertucci F, Finetti P, Rougemont J, et al. Gene expression profiling for molecular characterization of inflammatory breast cancer and prediction of response to chemotherapy.
Cancer Res
2004
;
64
:
8558
–65.
24
Bieche I, Lerebours F, Tozlu S, Espie M, Marty M, Lidereau R. Molecular profiling of inflammatory breast cancer: identification of a poor-prognosis gene expression signature.
Clin Cancer Res
2004
;
10
:
6789
–95.
25
Bardou VJ, Arpino G, Elledge RM, Osborne CK, Clark GM. Progesterone receptor status significantly improves outcome prediction over estrogen receptor status alone for adjuvant endocrine therapy in two large breast cancer databases.
J Clin Oncol
2003
;
21
:
1973
–9.
26
Pertschuk LP, Kim DS, Nayer K, et al. Immunocytochemical estrogen and progestin receptor assays in breast cancer with monoclonal antibodies. Histopathologic, demographic, and biochemical correlations and relationship to endocrine response and survival.
Cancer
1990
;
66
:
1663
–70.
27
Crowe JP, Hubay CA, Pearson OH, et al. Estrogen receptor status as a prognostic indicator for stage I breast cancer patients.
Breast Cancer Res Treat
1982
;
2
:
171
–6.
28
Allred DC, Clark GM, Tandon AK, et al. HER-2/neu in node-negative breast cancer: prognostic significance of overexpression influenced by the presence of in situ carcinoma.
J Clin Oncol
1992
;
10
:
599
–605.
29
Tandon AK, Clark GM, Chamness GC, Ullrich A, McGuire WL. HER-2/neu oncogene protein and prognosis in breast cancer.
J Clin Oncol
1989
;
7
:
1120
–8.
30
Van Laere S, Van der Auwera I, Van den Eynden GG, et al. Distinct molecular signature of inflammatory breast cancer by cDNA microarray analysis.
Breast Cancer Res Treat
2005
;
93
:
237
–46.