Gene expression profiles in breast tumors regarding the presence or absence of estrogen and progesterone receptors

Abstract

Estrogen acts via its receptor (ER) to stimulate cell growth and differentiation in the mammary gland. ER and progesterone receptor (PR), which is regulated by estrogen via ER, have been used as prognostic markers in clinical management of breast cancer patients. Patients with ER⁻ breast tumors have a poorer prognosis than patients with ER⁺ tumors. The aim of the present study was the identification of tumor-associated genes differentially expressed in breast tumors regarding the presence or absence of ER and PR hybridized with cDNA microarrays containing 4,500 tumor-derived expressed sequence tags generated using the ORESTES technique. Samples of human primary breast carcinomas from 38 patients were analyzed. The experiments were performed in triplicates and data from each element were acquired by phosphoimage scanning. Data acquisition was performed using the ArrayVision software. After normalization statistical analysis was applied. In a preliminary analysis, 98 differentially expressed transcripts were identified, 46 were found to be more expressed in ER⁺/PR⁺ and 52 were found to be more expressed in ER⁻/PR⁻ breast tumors. The biochemical functions of the genes in the reported expression profile are diverse and include metabolic enzymes, protein kinases, helicases, transcription factors, cell cycle regulators and apoptotic factors. ER⁻/PR⁻ breast tumors displayed increased levels of transcripts of genes associated with neurodegeneration and genes associated with proliferation were found in ER⁺/PR⁺ tumors. © 2004 Wiley-Liss, Inc.

Estrogens exert important roles in the development, differentiation and maintenance of several target tissues and are also associated with the development and growth of hormone-dependent tumors such as breast cancer.1

Most estrogen actions are thought to be mediated through its nuclear estrogen receptors, ERα and ERβ, which are members of the nuclear receptors superfamily that are ligand-induced transcription factors.2 The mechanism by which estrogen receptor (ER) mediates the transactivation of gene expression is complex.3 Besides the classical ligand-dependent mechanism of ER action in which the hormone-receptor complex regulates gene transcription through its interaction with ERE consensus DNA sequences, the ERs can also regulate gene transcription interacting with other promoter elements such as AP1,4 SP15 and CREs.6 Thus, estrogen receptors ERα and ERβ can transduce different hormonal signals depending on the ligand and the nature of the hormone-responsive element (HRE).2 The transactivation elicited by those receptors complexed with E2 may result in opposite signal transduction leading to opposite biologic response in the presence of AP1 and/or CRE sites.7, 8 In addition, several E2-responsive genes are regulated by DNA-independent or -dependent interactions of the ERα and SP1 proteins.9 Alternatively, the signal transduction by growth factors and their tyrosine kinase receptors, such as EGFR, IGFR, erbB-2 and other molecules like cAMP and dopamine, may lead to a ligand-independent ER activation, resulting from the phosphorylation of serine and tyrosine amino acid residues in the AF1 and AF2 domains in the estrogen receptor molecule.10 ER phosphorylation promotes receptor dimerization, association with coregulatory proteins and transcription transactivation.11

During the malignant transformation, the epithelial mammary cells can maintain the basal levels of ER or express high levels or loss of this receptor expression. The majority of breast cancers express ER, yet ER⁻ cancers comprise about 30% of all breast cancers.12 The origin of ER⁻ human breast cancer is unknown but it is possible that ER⁻ cancers may arise from a mammary epithelial cell that was ER⁻ during its normal phenotype or from an ER⁺ cell in which ER was transcriptionally downregulated.13

Patients with ER⁻ tumors show shorter disease-free interval and overall survival than patients with ER⁺ tumors.14, 15, 16 In addition, lymph node metastasis, considered to be the most important prognostic factor in breast cancer, does not alter these findings.17, 18 However, the ER expression is gradually lost during tumor progression, leading to a more undifferentiated state.19 The majority of ER⁺ tumors are diploid, well differentiated, more common in postmenopausal women and less aggressive than ER⁻ breast tumors.15, 20 However, the plethora of genes that might be associated with each tumor phenotype remains to be fully explored.

cDNA microarray technique allows the analysis of hundreds or thousands of genes simultaneously in a tissue or cell. This technique has been used to analyze differential gene expression in breast cancer cell lineages and in primary breast tumors to identify genes regulated by estrogen,21 tumor classifications and disease prognosis22, 23, 24, 25, 26, 27 and therapeutic response.28, 29

In the present study using cDNA microarrays containing 4,500 tumor-derived expressed sequence tags (ESTs), we have attempted to identify tumor-associated genes that can lead to a more precise classification of hormone-dependent and -independent primary breast tumors.

MATERIAL AND METHODS

Tissue samples

Samples of human primary breast carcinomas and adjacent normal tissue from 58 patients were obtained from A.C. Camargo Hospital (São Paulo, Brazil). Tumor samples were macrodissected to remove residual normal tissue and stored in liquid nitrogen. Only tumor samples with more than 60% of tumor cells were included in the present study. The largest diameter of the tumors was recorded. The number of lymph node metastases was determined by microscopic examination of an average of 24 lymph nodes per patient. Tumor metastases at lymph nodes were detected in 24 patients. All cases were submitted to a histopathologic review of tumor slides in order to confirm diagnosis. All tumors were classified according to the 1982 WHO Histological Typing of Breast Tumors. The clinical stage of the patients was determined according to the 1978 UICC TNM (tumor, nodes, metastases) staging system. Tumors studied were infiltrating ductal carcinomas. The median age of the patients at the time of diagnosis was 47.5 years (range, 25–75 years). Estrogen and progesterone receptor (PR) binding assays were performed by the classical dextran-coated charcoal method (DCC) as previously described.30

RNA extraction

Tissue specimens were pulverized under liquid nitrogen and total RNA was isolated according to the guanidine isothiocyanate method.31 The quality of the RNA samples was determined by 1% agarose gel electrophoresis and ethidium bromide staining. All RNA samples were treated with DNaseI for 30 min at 37°C to eliminate genomic DNA contamination.

cDNA microarrays, hybridization and data acquisition

A set of 4,500 tumor-derived EST clones generated by ORESTES32 was used to make cDNA microarray blots. Briefly, ORESTES clones were amplified by PCR using primers from sequences flanking the cloning site. The PCR products were purified using QIAquick 96 PCR purification kit (Qiagen, Chatsworth, CA), analyzed by 1.5% agarose gels and spotted onto nylon membranes robotically (GenomeSystem, St. Louis, MO). The ³³P-labeled cDNA probes were synthesized from 30 μg of total RNA treated with DNAseI using Superscript II reverse transcriptase (Life Technologies, Bethesda, MD) by oligo-dT-primed polymerization. The membranes were prehybridized for 40 min in 50% (V/V) phosphate solution (1 N Na⁺, 0.5 N PO4-3), 7% SDS, 1% BSA and 1 mM EDTA at 65°C. Hybridization was carried out for 16 hr using 50% (V/V) phosphate solution (1 N Na⁺, 0.5 N PO4-3), 7% SDS, 1% BSA and 1 mM EDTA at 65°C. The filters were washed twice for 30 min at 65°C in 50% phosphate solution, 1% SDS and 1 mM EDTA. The filters were then exposed to a phosphor-imaging screen for 6 hr to 2 days and scanned on the Storm Imager 860 (Molecular Dynamics, Sunnyvale, CA). The experiments were performed in triplicate. Samples were hybridized simultaneously against 2 nylon membranes on the same hybridization chamber. For tests purposes, a third hybridization has been done as a follow-up control but not considered for the analysis. Data acquisition was performed using the ArrayVision software (Molecular Dynamics), and normalization and statistic analysis were applied.

Data analysis

Data analysis was performed using R (http://cran.r-project.org), an open-source-interpreted computer language for statistical computation and graphics, and tools from the Bioconductor project (http://www.bioconductor.org), adapted to our needs. After image acquisition and quantification, saturated spots and spots with signal lower than or equal to background were identified and excluded from the analysis. Next, background-subtracted spot intensities were normalized by a global mean normalization procedure33, 34 and the final result was the average of duplicates. Next, we searched our data for differentially expressed genes in the ER⁺/PR⁺ and ER⁻/PR⁻ comparisons. We used a nonparametric test (nonpaired Wilcoxon or Mann-Whitney test) to determine the p-value for each individual gene without multiple-test correction. Next, we used Fisher's linear discrimination analysis to search for pairs and trios of genes such that data points representing signal intensities for all 3 genes for each sample would be well separated by a plane in the corresponding space. More precisely, for a given set of genes, this linear classification method searches for linear combinations of their expressions with large ratios between groups and within groups sum of squares.35 Along this direction where the ratio is largest, the 2 groups of data points, corresponding to ER⁺/PR⁺ and ER⁻/PR⁻ samples, are well separated by a line or a plane, depending whether we are considering pairs or trios of genes. This maximal ratio of sum of square, or its square root, which is denoted here by SVD, measures how well separated the 2 groups are.

Quantitative real-time PCR (qPCR)

qPCR was performed by using the GenAmp 5700 sequence detector (PE Applied Biosystems, Foster City, CA). cDNA was generated using the Prostar First Strand RT-PCR kit (Stratagene, La Jolla, CA). Each cDNA sample was analyzed in duplicate. PCR reactions were carried out in a total volume of 50 μl according to the manufacturer's instructions for SYBR Green PCR Core reagent (PE Applied Biosystems). The PCR primers used were as follows: Seladin-1 (DHCR24), forward primer 5′-GGTGCAGGACATGCTGGTGCC-3′, reverse primer 5′-CTCTGCCTCATTTCCTTTGGG-3′; amyloid β precursor (APP), forward primer 5′-ACCGCCGCCGCCTGGCCCTGG-3′, reverse primer 5′-GGACCGGATCTGAGCGGCTTTC-3′; heterologous nuclear ribonucleoprotein K (HNRPK), forward primer 5′-GTTGCCATCACCCACTGCAAC-3′, reverse primer 5′-CCTGCTAGACTCTGATGAA-3′; J domain containing protein 1 (JDP1), forward primer 5′-GGCGAAGGAGCCAGATGTCG-3′, reverse primer 5′-GTCTTGTCAGATTCTTCCAG-3′; glyceraldehyde-3-phosphate dehydrogenase (GAPDH), forward primer 5′-GAGCACCAGGTGGTCTCCT-3′, reverse primer 5′-TACCAGGAAATGAGCTTGACAAAG-3′. The relative expression was calculated by 2^−ΔΔCT, where CT = fluorescence threshold value; ΔCT = CT of the target gene − CT of the reference gene (GADPH); ΔΔCT = ΔCT of the tumor sample − ΔCT of the reference sample. The average value of all 4 normal tissue samples served as reference sample.

RESULTS

In this study aiming to identify tumor-associated genes related to breast cancer, we compared the pattern of gene expression between primary breast tumors with different estrogen and progesterone receptors status using cDNA microarrays. The status of the progesterone receptor was taken in consideration as an indication of the ER functionality. Total RNA isolated from 38 primary breast tumors (19 ER⁻/PR⁻ and 19 ER⁺/PR⁺) was hybridized with triplicates of a cDNA array containing 4,500 tumor-derived ESTs comprising 114 membranes. Data obtained from these membranes were normalized according to the methods described above. Two sets of 38 membranes (i.e., 76 membranes) were used in the analysis and a set of 38 membranes has been kept as a follow-up control. Data obtained from these 114 membranes indicate strong reproducibility. Indeed, the correlation coefficient between the 3 sets of 38 membranes (e.g., A, B and C) are 0.95 (A and B), 0.92 (A and C), 0.92 (B and C); as a comparison value, 0.77 is the correlation of ER⁺ and ER⁻ for all 114 membranes. The Wilcoxon test was applied to each cDNA spotted in the membranes to identify differentially expressed genes between the ER⁻/PR⁻ and ER⁺/PR⁺ breast tumors. A graphic representation of this analysis is shown in Figure 1. Each point in this volcano plot corresponds to a given transcript spotted into the membranes. The x-axis represents the difference between ER⁻/PR⁻ and ER⁺/PR⁺ tumor sample median log performed gene expressions and the y-axis corresponds to the p-values observed. The outlier's dots correspond to upregulated or downregulated genes between ER⁻/PR⁻ and ER⁺/PR⁺ breast tumors. The p-values obtained by the Wilcoxon test were used to rank the differentially expressed genes. Using this classification, we were able to identify a set of 98 cDNAs that could be good candidates for discriminating breast tumors regarding the steroid hormone receptor status. Forty-six of these transcripts were classified as more expressed in the ER⁺/PR⁺ breast tumors and 52 were classified as more expressed in the ER⁻/PR⁻ breast tumors. The genes in this expression profile are related to several biochemical and regulatory pathways. The differentially expressed genes are listed in Tables I and II. Increased expression levels of the Seladin-1 (DHCR24) gene transcripts were observed in ER⁻/PR⁻ breast tumor relative to the ER⁺/PR⁺ breast tumors. The expression of Seladin-1 gene was also analyzed, stratifying the breast cancer cases by other conventional prognostic factors, such as lymph node status and clinical stage. As shown in Figure 2, in addition to the negative correlation observed between Seladin-1 expression and steroid hormone receptor positivity, we found a positive correlation between the expression levels of Seladin-1 and advanced-stage tumors (stages III and IV).

Figure 1. Volcano plot. p-values from the nonparametric test versus the estimated ER⁻/PR⁻ and ER⁺/PR⁺ median difference.

Table I. List of the Transcripts with Significantly Higher Expression in ER⁺/PR⁺ Primary Breast Tumors

Accession #	Accession #	Description
HS.202179		EST
HS.348733	AF27304.8	CTCL tumor antigen se20-9 mRNA
NoM.6980		EST
Hs.260720	NM_021800	JDP1: J domain containing protein 1; DnaJ/HSP40 proteins
HS.249184	NM_007109	Transcription factor 19 (SC1) (TCF19)
Hs.129548	NM_002140	Heterogeneous nuclear ribonucleoprotein K (HNRPK)
HS.22350	AK27248	Hypothetical protein LOC56757 (FLJ23595)
Hs.171889	NM_020244	Choline phosphotransferase 1 (CHPT1)
HS.351575	NM_153029	Hypothetical protein FLJ31821
Hs.24817	NM_017684	Hypothetical protein FLJ20136
HS.146668	NM_020755	TDE1L: likely ortholog of mouse tumor differentially expressed 1
HS.344000	AK026100	Hypothetical protein FLJ22447
Hs.19054	NM_018530	Hypothetical protein PRO2521
Hs.23744	NM_024594	PANK3: pantothenate kinase 3
Hs.78921	NM_003488	AKAP1: A kinase (PRKA) anchor protein 1
Hs.380359	NM_005673	Solute carrier family 25 (SLC25A16)
Hs.12692	NM_017691	Hypothetical protein FLJ20156
Hs.183800	NM_002883	Ran GTPase activating protein 1 (RANGAP1; KIAA1835)
Hs.296847	NM_003119	SPG7: spastic paraplegia 7, paraplegin
HS.13011	NM_019096	GTP binding protein 2; putative GTP binding protein 2 (GTPBP2)
Hs.119257	NM_005231	EMS1: ems1 sequence
Hs.170160	NM_004761	RAB2L: RAB2, member RAS oncogene family-like, RGL2, KE1.5, HKE1.5
Hs.131905	NM_054020	CATSPER2: cation channel, sperm-associated 2
MGC14597	BE061374	EST similar to C20orf114: chromosome 20 open reading frame 114
Hs.321579	NM_021095	SLC5A6: solute carrier family 5 (sodium-dependent vitamin transporter), member 6
Hs.88556	NM_004964	HDAC1: histone deacetylase 1;HD1, RPD3, RPD3L1
Hs.172004	NM_003319	Titin (TTN)
HS_325530	NM_015219	EXO70: likely ortholog of mouse exocyst component protein 70 kDa homolog
HS.25682	NM_032478	MRPL38: mitochondrial ribosomal protein L38
Hs.234680	NM_013451	Fer-1-like 3, myoferlin (FER1L3)
Hs.107381	NM_018359	Hypothetical protein FLJ11200
HS.155313	NM_022105	DATF1: death-associated transcription factor 1
HS.31141	NT_022459	ROBO2: roundabout, axon guidance receptor, homolog 2 (Drosophila)
HS.109390	BF343288	EST Homo sapiens cDNA, 5′ end
Hs.6582	NM_015313	ARHGEF12: Rho guanine nucleotide exchange factor (GEF) 12
Hs.458148	NM_006013	RPL10: ribosomal protein L10
Hs.64096	NM_014772	Hypothetical protein KIAA0427
Hs.349106	AF113616	Mucin 3A, intestinal (MUC3)
HS.7149	AL834396	Formin homology 2 domain containing 2; (FHOD2) KIAA1902
Hs.75074	NM_004759	MAPKAPK2: mitogen-activated protein kinase-activated protein kinase 2
Hs.284162	NM_016304	C15orf15: chromosome 15 open reading frame 15
Hs.158203	NM_002313	ABLIM1: actin-binding LIM protein 1
HS.152925	AB033094	Hypothetical protein KIAA1268
Hs.7392	NM_024045	DDX50: DEAD (Asp-Glu-Ala-Asp) box polypeptide 50
Hs.194110	NM_02522	Hypothetical protein PRO2730
HS.3187	NM_002504	Nuclear transcription factor, X-box binding 1 (NFX1)

Table II. List of the Transcripts with Significantly Higher Expression in ER⁻/PR⁻ Primary Breast Tumors

Accession #	Accession #	Description
HS.319378		Homo sapiens transcribed sequence with moderate similarity to protein pir:I60307 (E. coli)
HS.155546	NM_015044	GGA2: golgi-associated, gamma adaptin ear containing, ARF-binding protein 2
Hs.367593	NM_052888	LOC114659: KIAA0563-related gene
Hs.198273	NM_005004	NADH dehydrogenase (ubiquinone) 1 beta (NDUFB8)
Hs.14732	NM_002395	Malic enzyme 1, NADP(+)-dependent, cytosolic (ME1)
Hs.177486	NM_000484	Amyloid beta (A4) precursor protein (APP)
Hs.3100	NM_005548	Lysyl-tRNA synthetase (KARS)
Hs.198427	NM_000189	Hexokinase 2 (HK2)
Hs.79356	NM_006762	Lysosomal-associated multispanning membrane protein-5 (LAPTM5)
Hs.135941	NM_014911	Adaptor-associated kinase 1 (AAK1) (KIAA1048)
Hs.23449	NM_018842	LOC55971:insulin receptor tyrosine kinase substrate
NoM.1249		EST
Hs.154672	NM_006636	Methylene tetrahydrofolate dehydrogenase (NAD^+-dependent) (MTHFD2)
Hs.74497	NM_004559	Nuclease-sensitive element binding protein 1 (NSEP1)
AF086762	NM_013279	C11orf9: chromosome 11 open reading frame 9 (KIAA0954)
Hs.278573	NM_000611	CD59: CD59 antigen p18–20
Hs.6879	NM_020188	DC13 protein (DC13)
Hs.79706	NM_000445	Plectin 1 (PLEC1)
Hs.79089	NM_006378	Semaphorin 4D (SEMA4D)
Hs.282410	NM_006888	Calmodulin 1 (phosphorylase kinase, delta) (CALM1)
Hs.3059	NM_016451	Coatomer protein complex, subunit beta (COPB)
NoM.7750		EST
HS.7736	NM_016504	Mitochondrial ribosomal protein L27 (MRPL27)
Hs.77290	NM_006755	Transaldolase 1 (TALDO1)
Hs279903	NM_005614	Ras homolog enriched in brain 2 (RHEB2)
HS.173108	NM_181784	SPRED2: sprouty-related, EVH1 domain containing 2
Hs.172207	NM_007363	NONO: non-POU domain containing, octamer-binding
Hs.153884	NM_005783	ATP-binding protein associated with cell differentiation (APACD)
Hs.158688	NM_015904	Translation initiation factor (IF2; KIAA0741)
Hs.81182	NM_006895	HNMT: histamine N-methyltransferase
Hs.75616	NM_014762	24-dehydrocholesterol reductase (DHCR24; SELADIN1)
HS.149377	NT_008705	Hypothetical protein DKFZp761L0424; KIAA1217
Hs.3260	NM_000021	Presenilin 1 (Alzheimer's disease 3) (PSEN1)
Hs.43857	NM_018837	SULF2: similar to glucosamine-6-sulfatases
Hs.75721	NM_005022	Profilin 1 (PFN1)
Hs.279784	NM_013388	Prolactin regulatory element binding (PREB)
Hs.4114	NM_005032	PLS3: plastin 3 (T isoform)
HS.288550	AK026809	Hypothetical protein FLJ23156
Hs.2877	NM_001793	CDH3: cadherin 3, type 1, P-cadherin (placental)
Hs.278562	NM_001307	Claudin 7 (CLDN7; claudin 9)
Hs.178710	NM_004859	CLTC: clathrin, heavy polypeptide (Hc)
Hs.16621	BE072346	Hypothetical protein DKFZP434I116 (KIAA1429)
Hs.3887	NM_002807	Proteasome (prosome, macropain) 26S subunit, non-ATPase, 1 (PSMD1)
HS.299269		Ye17d07.sl Homo sapiens cDNA, 3′ end (T93574)
Hs.70877	NM_015421	Hypothetical protein DKFZP564K2062
Hs.130740	NM_017758	Hypothetical protein FLJ20308
Hs.159263	NM_001849	Collagen, type VI, alpha 2 (COL6A2)
Hs.159479	NM_000512	GALNS: galactosamine (N-acetyl)-6-sulfate sulfatase
Hs.268555	NM_012255	XRN2: 5′–3′ exoribonuclease 2
Hs.21321	NM_145808	Granule cell differentiation protein (FLJ31098)
Hs.343564	NM_003193	TBCE: tubulin-specific chaperone
Hs.433222	NM_006432	NPC2: Niemann-Pick disease, type C2

Figure 2. Seladin-1 expression according to steroid hormone receptor status, lymph node status and clinical stage. The box plot shows the data distributed in relation to the median values for each patient stratified by ER/PR status, nodes and stage. Lower and upper bars linked to each box indicate the borders of the normal distribution; lower and upper box edges represent the first and third quartile within each distribution; the bar within each box provides the Wilcoxon variability about the median (steroid hormone receptor status, p = 0.03; lymph node status, p = 0.60; and clinical stage, p = 0.03; +, indicates data outliers). Negative, ER⁻/PR⁻ tumor samples; positive, ER⁺/PR⁺ tumor samples; nodes 1, absence of lymph node involvement; nodes 2, presence of lymph node involvement; stage 1, early clinical stage tumors (stages I and II); stage 2, advanced clinical stage tumors (stages III and IV).

After selecting the 98 differentially expressed cDNA clones, we used a supervised analysis (described above) and made an exhaustive search in order to find pairs and trios of genes whose pattern of expression in individual arrays allows a good linear separation between breast tumors with different steroid hormone receptor status. We found some pairs of genes showing this potential. A linear correlation between the expression of the JDP1 and HNRPK gene transcripts was able to make a good separation between the ER⁺/PR⁺ and ER⁻/PR⁻ breast tumors (Fig. 3).

Figure 3. Scatter plot showing the linear correlation between the expression of HNRPK and JDP1 genes. Each point on the plane represents the normalized log expression intensities for each tumor sample. •, ER⁻/PR⁻ breast tumor; ▴, ER⁺/PR⁺ breast tumors.

The transcripts of the Seladin-1, APP, HNRPK and JDP1 genes were further validated in a series of 20 primary breast tumors (ER⁺/PR⁺, 10 cases; ER⁻/PR⁻, 10 cases) and 4 normal breast tissue samples by qPCR (Fig. 4). Overall, the expression patterns of these genes observed by qPCR analysis were consistent with the results generated by the microarray method.

Figure 4. Expression analysis of the Seladin-1, APP, HNRPK and JDP1 genes in primary breast tumors. The relative expression was measured by qPCR, normalized to GAPDH as the reference gene. Height of the bars represents the relative gene expression for individual tumors taking the normal breast tissue as reference samples. (a) Seladin-1 and APP. (b) HNRPK and JDP1.

To investigate the potential regulation of the HNRPK and JDP1 by estrogen, using the BLAT tool we searched for 5′ flanking sequences that might act as promoter region for these genes. No potential promoter region could be found for the JDP1 gene. A CpG-rich sequence of 1,815 bp located 5′ upstream of the HNRPK gene was identified as a potential promoter region for this gene (Fig. 5). A search for potential transcription factor-binding sites was done using publicly available tools.36 This promoter region contains motif sequences for several transcription factors, including AP1, SP1 and CREB. In addition, this region also contain several widely spaced EREs and one perfect palindromic ERE.

Figure 5. Nucleotide sequence of the putative 5′ flanking region of the HNRPK gene. The sequence data were obtained from www.genome.uscs.edu. Numbers beside the sequence indicate the nucleotide position in relation to the transcription start site (+1). The exons 1 of the HNRPK gene is represented in capitals and the noncoding sequences are represented in small cases. The consensus sequences for regulatory elements and transcription factor binding sites are indicated in bold and underlined.

DISCUSSION

A large proportion of the differentially expressed transcripts identified here are new genes, represented by anonymous ESTs (without RefSeq) or hypothetical proteins. Several known genes associated with different biologic pathways already known to be associated with the tumorigenic process were also identified as differentially expressed in the breast tumors stratified by steroid hormone receptors status. Among the known genes are those related to the transduction of mitogenic signal, cellular adhesion and motility, metabolic enzymes, protein kinases, several transcription factors and apoptotic factors. Due to the diversity and complexity of the expression profile observed, we will focus our discussion on a limited number of genes.

The gene expression profile displayed by ER⁻/PR⁻ breast tumors included increased transcript levels of genes that act as apoptotic and antiapoptotic factors in cells of the nervous central system and are associated with neurodegeneration, such as Seladin-1, APP and PSEN1. Seladin-1 is a new gene and its physiologic role in breast epithelium is unknown. The Seladin-1 gene (also named KIAA0018 and DHCR24, 24-dehydrocholesterol reductase) contains 9 exons spanning approximately 46 Kb of genomic DNA on chromosome 1p33-p31.1.37, 38, 39 Previous data from our group identified Seladin-1 gene transcripts in ER⁻/PR⁻ primary breast tumors using differential display reverse transcriptase (DDRT)-PCR.40 In the present study, we have also observed that advanced-stage breast tumors (stages III and IV) showed higher levels of Seladin-1 gene expression compared to early-stage breast tumors (stages I and II), possibly reflecting its association with a more aggressive nature. Seladin-1 encodes for a protein with homology with a family of FAD-dependent oxireductases, which play a role in cholesterol biosynthesis.39 In neuroglioma cells, Seladin-1 acts as a antiapoptotic factor inhibiting caspase 3 in cells overexposed to toxicity mediated by amyloid β accumulation or by oxidative stress.38 The overexpression of Seladin-1 may function as a protective factor against the toxicity elicited by the amyloid β accumulation and its downregulation might be associated to an increased susceptibility to Alzheimer's disease (AD) development.38 The APP (amyloid β precursor) and PSEN1 are also associated with AD. The biologic function of the Presenilin-1 has not been fully established yet,41 but this protein appears to play an important role in APP processing and mutations in the PSEN1 gene have been associated with 18–50% of the familial AD cases.42

The role played by estrogen in the transcription regulation of Seladin-1 and Presenilin-1 is not known yet, but the regulation of the APP gene has been attributed to the estrogen action. Epidemiologic studies have shown an association between the reduction of estrogen levels and the development of AD, which represents one of the most common neurodegenerative disease in postmenopausal women.43 Estrogen replacement therapy has been associated with a reduction in the relative risk for AD development following menopause.43 In addition, experimental studies have shown that estrogen decreases the levels of APP expression, alters APP processing, decreases amyloid β production and the apoptosis induced by amyloid β in neuronal cells.44, 45 Our data suggest that estrogen receptors might be associated with the repression of the transcriptional regulation of Seladin-1, APP and PSEN1; however, experimental studies are needed to address that. Investigating the effects of the BRCA1 expression in the transcriptome of EcR-293 cells in culture, Welsch et al.46 identified APP as being induced by the BRCA1 expression. High levels of BRCA1 expression inhibit ERα transcriptional activity.47 These data suggest that the upregulation of APP transcripts by BRCA1 might be a result of the inhibition of ER transactivation activity.

Using a supervised analysis, we found pairs of genes whose pattern of expression would be useful as markers to better identify the hormone-responsive phenotype. Among those we identified a pair of genes composed by JDP1 (J domain containing protein 1; DnaJ/Hsp40) and HNRPK (heterologous nuclear ribonucleoprotein K; hnRNP K) that appeared to be particularly interesting. The JDP1 and HNRPK transcripts were found to be more expressed in ER⁺/PR⁺ breast tumors.

The JDP1 gene encodes a DnaJ/Hsp40 protein, which is a member of the small-molecular-weight heat shock protein family. DnaJ/Hsp40 proteins are expressed in several tissues and acts as a cochaperone regulating protein folding, transport, translation initiation and gene expression.48 Hsp40 proteins coregulate the activity of Hsp70 proteins in an ATP-dependent manner, preventing the accumulation of improperly folded proteins and facilitating their degradation via ubiquitin.48, 49 Schneider et al.50 provide evidence that the eukaryotic DnaJ homologue can interact with the major cytoplasmatic molecular chaperone Hsp90, which interacts with several proteins, including steroid hormone receptors, such as ER, preventing receptor activity in the absence of the hormone.51 In yeast, Ydj1/Hsp40 mutants displayed high basal levels of ER activity in the absence of estrogen, suggesting that Ydj1/Hsp40 is important for the receptor regulation by the Hsp90 folding pathway.52 These data suggest that the cochaperone activity of the DnaJ/Hsp40 in the Hsp90 pathway might also be important for the integrity and functionality of ER in the mammary normal gland and breast tumor tissue.

The HNRPK gene encodes for the hnRNP K protein initially described as a member of the heterologous ribonucleoproteins family that bind nascent pre-mRNA and participate in the mRNA processing and transportation to the cytoplasmic compartment.53 hnRNP K was also characterized as a transcription factor that can bind DNA double strand at pyrimidine-rich motifs (CCCTCCCCA; CT element), interact with TATA-binding proteins (TBPs) and regulate gene transcription of several genes, such as c-myc54 and SRC.55 As the expression of JDP1 and HNRPK transcripts showed a positive linear correlation associated with the positivity for estrogen and progesterone receptors, we speculate about the possible estrogen responsiveness of those genes. We identified a putative promoter region 5′ upstream of the HNRPK gene that contains several transcription factor motifs, including many SP1 sites and half ERE consensus sequences and a perfect palindromic ERE, suggesting that ER might play an important role in the transcriptional regulation of this gene. HNRPK transcript upregulation via ER could occur by a hormone-dependent or -independent mechanism. On the other hand, the transactivation of HNRPK might be a result of the cellular cross-talk between ER and growth factor signaling pathways.10

Increased levels of HNRPK transcripts were observed in MCF-7 breast cancer cells in the presence of EGF or HRG (heregulin-β4).56 Mandal et al.56 also demonstrate that the expression of hnRNP K significantly increases c-myc expression and the proliferation rate of MCF-7 cells. In addition, they observed increased expression levels of hnRNP K in grade III primary breast tumors, suggesting that the expression of HNRPK might be associated with enhanced growth rate of breast cancer cells.

In summary, we have identified different sets of genes whose transcripts are differentially expressed in breast tumors regarding the estrogen and progesterone receptor status. Our data could provide new information in estrogen signaling and tumor growth. Additional studies are required to determine the functional and clinical significance of these genes in breast cancer development and its potential to be used as tumor biomarkers.

Acknowledgements

The authors thank Dr. Luis Fernando Reis (Ludwig Institute for Cancer Research São Paulo Branch) for kindly providing the cDNA arrays.