Genomic profiling of C/EBPβ2 transformed mammary epithelial cells

Abstract

C/EBPβ is essential for mammary gland growth and development and has been associated with poor prognosis in breast cancer. Overexpression of C/EBPβ2 in MCF10A cells results in a variety of cancer phenotypes including EMT and ErbB independence. IL1β is dramatically upregulated in MCF10A-C/EBPβ2 cells but there is little, if any, processing to the mature 17 kD form. Although proIL1b has previously been considered to be biologically inactive, we demonstrate proIL1b is not only localized to the nucleus, but is also tightly associated with the chromatin. We show that proIL1β is bound at specific locations in the genome and is positioned in such a way to play a role in the cancer phenotypes observed in MCF10A-C/EBPβ2 cells. Moreover, nuclear IL1β is detected in some human breast tumor samples. This study demonstrates the presence of nuclear proIL1β in transformed mammary epithelial cells providing the first evidence that IL1β may be a dual function cytokine.

Introduction

Inflammation contributes to the development and progression of breast cancer. Interleukin-1β (IL1β) is a key mediator of inflammation. IL1β is synthesized as a 31 kD proform that is cleaved by caspase-1 to a mature 17 kD secreted form. Only the mature form of IL1β binds to the interleukin 1 receptor 1 (IL1R1) and for this reason, proIL1β is considered biologically inactive. Many studies highlight the importance of IL1β in cancer, including breast cancer. Secreted IL1β plays a key role in carcinogenesis, tumor invasiveness and tumor-host interactions.^1,2 IL1β is expressed by both the malignant cells as well as the microenviroment in breast cancer.^3–5 In fact, IL1β has been demonstrated to be present in 90% of invasive breast carcinomas by ELISA and correlates with high tumor grade.³

Although IL1β is thought to be active only when cleaved into the 17 kD form and subsequently secreted, several other cytokines have been shown to have intracellular nuclear functions in their proform. These dual function cytokines include two members of the interleukin-1 family, IL1α and IL33, and the highmobility group box 1 protein (HMGB1). In the present study, we demonstrate proIL1β is upregulated by the transcription factor CCAAT/enhancer binding protein-b (C/EBPβ)-2 in mammary epithelial cells. C/EBPβ is a basic-leucine zipper transcription factor essential for mammary gland development.^6,7 C/EBPβ also plays a critical role in Ras transformation, as recently shown by the conditional elimination of C/EBPβ in keratinocytes which conferred complete resistance of the skin to chemical carcinogenesis involving mutant Ras.⁸ Histological studies indicate C/EBPβ is involved in the progression of human breast cancer and is predictive of poor prognosis.⁹ Although C/EBPβ is intronless, three different proteins can be made from its mRNA using alternative translation initiation sites.¹⁰ The first two isoforms, C/EBPβ1 and C/EBPβ2, are transcriptional transactivators; C/EBPβ3 lacks the transactivation domain and functions as a repressor. Although different by only 23 N-terminal amino acids, C/EBPβ1 and C/EBPβ2 are not biologically equivalent.^11–13 Importantly, C/EBPβ2 is expressed at high levels in 70% of invasive mammary carcinomas but is not detected in normal tissue obtained from reduction mammoplasties.¹² Additionally C/EBPβ2 overexpression in MCF10A cells, an immortalized but not transformed mammary epithelial cell line, results in multiple cancer phenotypes. These phenotypes include anchorage independence, an invasive phenotype, epidermal growth factor (EGF) independence, altered acinar architecture in 3D-culture models and epithelial to mesenchymal transition (EMT).^14,15

To gain insight into the cancer phenotypes acquired by MCF10A-C/EBPβ2 cells, we characterized global changes in gene expression. C/EBPβ2 overexpression in MCF10A cells dramatically upregulated (more than 30 fold) IL1β, IL1 receptor 2 (IL1R2), a decoy receptor for IL1β that lacks the cytoplasmic signaling domain and IL1 receptor like 1 (IL1RL1) whose cognate ligand is IL33 (a member of the IL1 family). IL1β is a classic target gene for C/EBPβ and chromatin immunoprecipitation (ChIP) assays confirmed C/EBPβ2 is bound to the IL1β promoter. Surprisingly, IL1β present in MCF10A-C/EBPβ2 cells is not cleaved and is found in the nucleus, tightly associated with chromatin. After completing a genome wide location analysis of proIL1β binding sites in chromatin, we used multiple bioinformatic approaches to categorize genes in the vicinity of nuclear proIL1β whose chromatin environment could be influenced if nuclear proIL1β functions via chromatin remodelling like the other dual function cytokines.

Results

Genomic profiling reveals C/EBPβ2 gene regulation in MCF10A cells.

MCF10A cells are immortalized, but not transformed, mammary epithelial cells that require EGF for growth. Overexpression of C/EBPβ2 in these cells results in multiple cancer phenotypes including EGF independent growth.^14,15 We characterized C/EBPβ2 regulation of gene expression in MCF10A cells using Affymetrix human genome U133 Plus 2.0 microarrays. 443 genes were found to be statistically differentially expressed upon C/EBPβ2 overexpression. Of these differentially expressed genes, 86 were found to be upregulated 2 fold or more, while 121 where found to be downregulated 2 fold or more (Supplemental File 1). C/EBPβ2 overexpression in MCF10A cells results in EMT and genomic profiling revealed upregulation of multiple genes (SPARC, CYR61, CTSB, CDH1) important in EMT and correlated with poor prognosis in human breast cancer (Table 1).¹⁵ Our previous work indicated ErbB family members were not responsible for the EGF independence observed in MCF10A-C/EBPβ2 cells based upon the observations that ErbB1 and ErbB2 are not activated nor is the Ras/Raf/ERK1/2 pathway activated in MCF10A-C/EBPβ2 cells.¹⁵ Genomic profiling confirmed the ErbB family of receptors and their ligands are not upregulated (Table 1). Bioinformatics methods to determine pathway involvement based on gene profiling data, such as PathwayAssist analysis, also did not detect known ErbB pathways, but did identify EMT related pathways involving regulation of the actin cytoskeleton (p value 0.035) and focal adhesions (p value 0.040).

Table 1.

Key results of genomic profiling in C/EBPβ2 overexpressing MCF10A cells

	Gene symbol	Fold change	p value
	SPARC	6.9	0.033
EMT related genes	CYR61	6.0	0.027
	CTSB	2.5	0.009
	CDH1	−5.6	0.008
	EGFR	1	n/a
	ERBB2	N.E.	n/a
	ERBB3	N.E.	n/a
	ERBB4	N.E.	n/a
	EGF	N.E.	n/a
	AR	N.E.	n/a
Erbβ Signaling	TGFA	N.E.	n/a
	BTC	N.E.	n/a
	HBEGF	N.E.	n/a
	EPR	N.E.	n/a
	HRG	N.E.	n/a
	NRG2	N.E.	n/a
	NRG3	N.E.	n/a
	NRG4	N.E.	n/a
	IL1R2	315.7	0.002
Interleukin-1 family	IIRL1	50.3	0.048
	IL1β	31.9	0.002

C/EBPβ2 regulates key genes known to be involved in EMT but does not alter expression of known Erbb signaling components. Strikingly, C/EBPβ2 overexpression in MCF10A cells dramatically upregulates three interleukin-1 family members. Changes in expression for bolded genes have been confirmed using real-time PCR. No expression (N.E.) is used as the fold change for genes whose expression was not detected at a significant level.

C/EBPβ2 overexpression in MCF10A cells leads to the dramatic upregulation of three interleukin-1 family members.

C/EBPβ2 overexpression in MCF10A cells dramatically upregulated (30 fold or more) several members of the interleukin-1 family (Table 1) including interleukin-1β (IL1β), interleukin-1 receptor 2 (IL1R2) and interleukin-1 receptor like 1 (IL1RL1, also called ST2). These results were confirmed using real-time PCR. To determine if the three highly upregulated interleukin-1 family genes are direct targets of C/EBPβ2, we used chromatin immunoprecipitation (ChIP). To examine only C/EBPβ2 and not other isoforms of C/EBPβ, an antibody to the T7 epitope was used to immunoprecipitate C/EBPβ2 bound regions of the genome followed by qPCR. As seen in Figure 1, C/EBPβ2 binds directly to the IL1β promoter at two distinct locations, −1 to −173 and −2,258 to −2,805 base pairs upstream of the IL1β transcription start site. Previous studies have shown C/EBPβ binds the IL1β promoter at −17 and −107 and also −168 and −258.¹⁶ We thus identify a new binding site further upstream and demonstrate for the first time that the C/EBPβ2 isoform can bind to the IL1β promoter. ChIP was also used to investigate 5,000 base pairs upstream and downstream of the transcriptional start site of IL1R2 and IL1RL1, but C/EBPβ2 was not found to bind anywhere within these regions, suggesting either C/EBPβ2 regulation is indirect or accomplished over a longer distance.

Figure 1.

C/EBPβ2 directly binds the IL1β promoter but not the IL1R2 or IL1RL1 promoters. Chromatin immunoprecipitation (ChIP) was performed using an antibody to the T7 epitope to pull-down C/EBPβ2 bound regions of the genome. Shown are the quantitative real-time PCR (Q-PCR) analysis for the negative controls (A) and the promoter regions of (B) IL1β. For negative controls three untranscribed regions of the genome were used. Q-PCRs were run in triplicate and the averaged Ct values were transferred into copy numbers of DNA using a standard curve of genomic DNA with known copy numbers. Results are normalized for primer pair amplification efficiency using the Q-PCR value obtained with unprecipitated genomic DNA (Input DNA). Results are presented as Binding Events Per 1000 Cells for the promoter region tested. Error bars correspond to the standard deviations from triplicate Q-PCR reactions.

C/EBPβ2 overexpression in MCF10A cells regulates IL1β and IL1R2 protein expression.

We performed immunoblot analysis to determine the protein levels of C/EBPβ2, IL1β, IL1R2 and IL1RL1 (Fig. 2A–C). IL1R2 and IL1β were detected at the protein level in MCF10A-C/EBPβ2 cells, but IL1RL1 was not. Interestingly, only proIL1β (31 kD) was detected in MCF10A-C/EBPβ2 cells (Fig. 2C). The active 17 kD form of IL1β was not detected by immunoblot analysis in either whole cell lysates or conditioned media (Fig. 2C and data not shown). However, low levels (5–50 pg/mL) of IL1β are known to be biologically active. Therefore, an enzyme-linked immunosorbent assay (ELISA) was performed on conditioned media. We did not detect IL1β in MCF10A cells, but conditioned media from MCF10A-C/EBPβ2 cells had 18.0 ± 0.3 pg/mL (Fig. 2D). Nonetheless, addition of IL1β up to a hundred fold higher amount than was detected by ELISA did not enable MCF10A cells to grow in the absence of EGF (Fig. 2E) or acquire any other hallmarks of transformation (data not shown). We also blocked IL1 signaling in MCF10A-C/EBPβ2 or MCF10A cells using the recombinant IL1 receptor antagonist (IL1ra). IL1ra binds interleukin-1 receptor 1 (IL1R1), and prevents IL1β mediated signal transduction.¹⁸ Even at 1,000 fold molar excess, IL1ra had no effect on the growth of MCF10A-C/EBPβ2 cells, either in the presence or absence of EGF (Fig. 2F). Likewise, IL1ra did not affect the growth of MCF10A cells in the presence of EGF. IL1ra was not tested on MCF10A cells in the absence of EGF because these cells are unable to grow without EGF. Taken together, ‘classical’ IL1β signaling does not appear to play a role in the EGF independence of MCF10A-C/EBPβ2 cells. This is consistent with the fact that MCF10A-C/EBPβ2 cells express high levels of the IL1R2 decoy receptor which would be expected to block signaling by a small amount of secreted IL1β.

Figure 2.

Analysis of IL1β and IL1R2 at the protein level. Equal amounts of protein extracts from MCF10A cells or MCF10A cells overexpressing C/EBPβ2 were subjected to immunoblot analysis with (A) anti-T7, (B) anti-IL1R2 or (C) anti-IL1β. ELISA was also performed to determine the concentration of IL1β in the culture supernatant of these cells (D). All of these experiments were repeated at least three times with similar results. (E) MCF10A cells were cultured in the presence or absence of EGF and various levels of recombinant human 17 kD IL1β (R&D systems) as indicated. (F) MCF10A cells or MCF10A-C/EBPβ2 cells were cultured for 48 hrs in increasing amounts of IL-1ra (Kineret; Biovitrum). MCF10A-C/EBPβ2 cells were also cultured in increasing amounts of IL-1ra in the absence of EGF. Cell counts are the average of triplicate determinations with error bars indicating standard deviation from the mean.

ProIL1β present in MCF10A-C/EBPβ2 cells is localized to the nucleus and is tightly associated with the chromatin.

Although proIL1β is thought to be inactive, a growing number of cytokines have been shown to play nuclear roles in their proform. We ascertained the localization of IL1β in MCF10A-C/EBPβ2 cells by indirect immunofluorescence (Fig. 3A). IL1β was detected in the nucleus of MCF10A-C/EBPβ2 cells. Visual inspection of Figure 3A indicates some cells express higher levels of IL1β and there is also some cytoplasmic IL1β in some of the cells. This may be due to the variable expression of C/EBPβ2 within the population, given a mix of retroviral integration sites within the population. Performing the same indirect immunofluorescence in MCF10A cells results in no signal, as expected and staining using the secondary antibody alone on the MCF10A-C/EBPβ2 cells gives no background (Fig. 3A and data not shown). MCF10A and MCF10A-C/EBPβ2 cells were also stained using DAPI to visualize the cell nuclei. The DAPI stain has been false colored red. Merging the IL1β and DAPI demonstrates that the staining is colocalized. This colocalization confirms IL1β is nuclear in the cells. Nuclear localization of proIL1β was confirmed in MCF10A-C/EBPβ2 cells using cell fractionation (Fig. 3C). The proform of IL1β is detected in both the cytoplasm and nucleus, whereas the 17 kD cleaved form of IL1β is not detected in the MCF10A-C/EBPβ2 cells (although the use of recombinant 17 kD IL1β demonstrates that the antibody used in this experiment would detect 17 kD IL1β if it were present). These results demonstrate a significant proportion of proIL1β is present in the nucleus of MCF10A cells overexpressing C/EBPβ2.

Figure 3.

ProIL1β is localized to the nucleus. (A) Indirect immunofluorescence using anti-IL1β was performed on MCF10A-C/EBPβ2 cells and parental MCF10A cells. DAP I was used to stain the nuclei (and is false colored red to allow for colocalization with IL1β)(B). Nuclear localization of proIL1β was also assessed by cellular fractionation of MCF10A (B, lanes 2–4) or MCF10A-C/EBPβ2 cells (B, lanes 5–7) followed by immunoblot analysis. Recombinant purified human 17 kD IL1β was analyzed in lane 1 as a positive control. Efficient cellular fractionation was confirmed using anti-TBF as a nuclear specific control and anti-β tubulin as a cytoplasmic marker.

Nuclear extraction in buffers of increasing ionic strength can be used to determine how tightly a protein is associated with the chromatin. The first extraction (0.1 M NaCl) removes only those proteins that are very loosely associated with the chromatin. Extraction in 0.3 M NaCl or 0.6 M NaCl releases a large number of nuclear proteins, including a variety of transcription factors. For example, the well-characterized sequence-specific basic leucine zipper transcription factor, CREB (cAMP response element binding protein), is completely extracted from nuclei by 0.6 M NaCl (Fig. 4, lane 5). The core histones are among the most tightly bound proteins in chromatin. A 1 M NaCl extraction removes 50–75% of histones H2A and H2B (see H2A, Fig. 4, lane 6) whereas histones H3 and H4 require 2 M NaCl for complete extraction. Importantly, although some proIL1β appears loosely associated and is extracted from nuclei by 0.1 M to 0.3 M NaCl, the majority of nuclear proIL1β is tightly bound to chromatin, resisting extraction by 1.0 M NaCl (Fig. 4, lane 7). These results demonstrate proIL1β is not only present in the nucleus but also interacts with the chromatin even more tightly than the core histone H2A.

Figure 4.

ProIL1β is tightly associated with the chromatin. Nuclei from MCF10A cells overexpressing C/EBPβ2 were extracted into five fractions sequentially: buffer B containing 0.1 M NaCl (lane 3); or 0.3 M NaCl), (lane 4); or 0.6 M NaCl (lane 5); or 1.0 M NaCl (lane 6) and the nuclear pellet (lane 7). Lane 2 is the cytoplasmic fraction and lane 1 contains whole cell extract. Cellular fractionation was monitored using anti-CREB or anti-H2A as nuclear markers and anti-β tubulin as a cytoplasmic marker.

ChIP-chip demonstrates nuclear proform IL1β binds specific regions of the chromosome.

To gain further insight into nuclear proIL1β, we performed chromatin immunoprecipitation followed by hybridization to the Human Promoter 1.0R array (Affymetrix) to identify proIL1β binding sites in chromatin across the genome. Chromatin bound by proIL1β was collected by immunoprecipitation with the IL1β monoclonal antibody MAB201 (R&D Systems), the same antibody used in previous figures. ProIL1β was found to bind at 204 locations (Fig. 5) and all chromosomes contained at least one proIL1β binding site, but these sites did not associate randomly across the chromosomes. Nearly two thirds of the sites were found in clusters of two or more closely spaced sites with several of the highly statistically significant clusters containing five to seven proIL1β sites (seen visually in Fig. 5 and listed with p values in Suppl. File 2). We are currently uncertain as to the functional significance of these multiple site clusters except to note that one of them (7 sites clustered on chromosome 6) coincides with a locus highly enriched for MHC class II genes. Combining the ChIP-chip analysis of proIL1β with the Affymetrix expression profiling of MCF10A-C/EBPβ2 cells, we can conclude that the sites of proIL1β binding are not closely associated with the promoters of the 443 genes found to be differentially expressed upon C/EBPβ2 overexpression. Therefore, it seems unlikely that pro-IL1β is functioning as a typical trans-acting transcription factor.

Figure 5.

ProIL1β binds at distinct chromosomal locations. Chromatin immunoprecipitation (ChIP) was performed using an antibody to IL1β to pull-down proIL1β bound regions of the genome. After purification, precipitated DNA was then hybridized to the Human Promoter 1.0R array (Affymetrix). Shown are the 204 sites (multicolor) detected to be 2-fold or more enriched when IL1β ChIP DNA was hybridized to the array compared to non ChIP DNA. Approximately two thirds of these sites bound in clusters. ProIL1β binding was highly correlated with metastasis initiation genes shown here by black bars.

The dual function cytokines identified to date mediate changes in the chromatin landscape altering the accessibility to transcription factors of a coordinated set of genes.^18–22 In the event that proIL1β may regulate the chromatin landscape, we used bioinformatics tools to categorize the genes in the “neighborhood” of proIL1β binding since these genes would be the most likely to be influenced by any proIL1β-dependent changes in chromatin conformation. Pathway analysis using PANTHER for the set of genes within 500 KB of proIL1β binding sites identified multiple pathways known to be affected in cancer (Table 2). These include the following signaling pathways: integrin, EGF, PDGF, Ras and apoptosis. In a similar approach, we performed over-representation analysis to determine if pro1L1b binding was correlated with genes involved in tumor initiation, metastasis initiation, metastasis progression and/or metastasis virulence using gene sets defined in a recent review by Chiang and Massague.²³ Interestingly, proIL1β binding is highly significantly correlated with all of these pathways (Table 2). Most highly significant (1.00E-28) were eleven genes (RhoC, LOX, VEGF1,2, CSF-1, ID1, TWIST, MET, FGFR, MMP-2, NEDD9) associated with metastasis initiation. As shown in Figure 5, six of the eleven genes were located in the neighborhood (<750 kB) of proIL1β binding sites, with one gene (VEGFB) in close proximity (<32 kB) and another (VEGFC) harboring a site within the 3′ end of the gene. These data indicate that future studies to define the mechanism of proIL1β binding to chromatin and its precise functional consequences are warranted.

Table 2.

Bioinformatic analysis connects chromatin associated proIL1β with multiple pathways known to be altered in cancer and with genes known to be involved in tumor initiation and metastasis

Over-representation analysis using panther analysis to determine altered signaling pathways
Signaling pathway	Found	Expected	p value
Integrin	38	19.94	0.031
PDGF	33	16.6	0.039
EGF receptor	27	13.18	0.087
Ras	19	7.99	0.104
Apoptosis	24	11.51	0.137

Over-representation analysis looking at known genes involved in specific tumor stages
	Total genes	ProIL1β associated	p value
Tumor initiation	9	3	7.22E-12
Metastatic initiation	11	7	1.00E-28
Metastatic progression	4	2	1.17E-08
Metastatic virulence	5	1	0.000222

Over-representation analysis performed and statistical significance determined using Panther Analysis reveals increased expression of genes in multiple cancer related pathways. In addition, over-representation analysis performed and statistical significance determined using modified formula from Backes et al. indicates IL1β is significantly correlated with genes important for tumor and metastatic initiation, metastatic progression and metastatic virulence.³⁸ See Methods for detailed description.

Nuclear IL1β is present in human breast cancer samples.

To determine if nuclear IL1β was present in human breast tumors we performed immunohistochemical analysis on samples of invasive ductal carcinoma. Nuclear IL1β was detected in both the cancer cells themselves and also in infiltrating leukocytes (Fig. 6A, C and E, magnifications in B, D and F). Previous studies have detected IL1β in 90% of invasive ductal carcinoma via ELISA.³ The samples displaying nuclear IL1β shown in Figure 6 are both HER2 positive (A, E) and negative (C). Although the sample shown in C is PR positive, all three samples happen to be ER negative. We conclude that nuclear IL1β is present in tumors from some breast cancer patients, although a larger study will be necessary to determine if nuclear IL1β correlates with tumor subtype, stage and/or grade.

Figure 6.

Nuclear IL1β in human breast cancer samples. The tumors shown have the following characteristics. (A) Stage II–III, T2N0M0, ER⁻, PR⁻ and HER2⁺⁺⁺. (C) Stage II–III, T2N0M0, ER⁻, PR⁺⁺ and HER2⁻. (E) Stage III, T4N3MX, ER⁻, PR⁻ and HER2⁺⁺. Magnification of each of the tumors is shown in (B, D and E).

Discussion

C/EBPβ2 overexpression in MCF10A cells results in the upregulation of IL1β, IL1R2 and IL1RL1 all of which are known to be upregulated in human breast cancer.^3,4,24 We demonstrated by chromatin immunoprecipitation that IL1β is a direct target of C/EBPβ2 transactvation in MCF10A cells, showing for the first time that this particular isoform of C/EBPβ can bind the promoter. It is well established that secreted IL1β plays an important role in breast cancer.²⁵ Although MCF10A-C/EBPβ2 cells have a small amount of IL1β in the culture medium, the IL1R2 protein is also concomitantly elevated in these cells. As a decoy receptor devoid of any cytoplasmic signaling domain, IL1R2 would be expected to block mature IL1β from classically signaling though IL1R1. Indeed, we found that addition of interleukin-1 receptor antagonist did not alter the EGF independent growth of MCF10A-C/EBPβ2 cells nor was the addition of exogenous mature IL1β to the culture medium sufficient for ErbB independent growth.

While functional studies of IL1β have always focused on the mature, secreted protein because proIL1β is considered biologically inactive, our results presented here clearly call this notion into question. The majority of IL1β protein present in MCF10A-C/EBPβ2 cells is found in the 31 kD proform and a substantial fraction of proIL1β is very tightly bound to chromatin in the nucleus of these cells. Therefore, it seems probable that IL1β, like IL1α, IL33 and HMGB1, is also a dual function protein, with roles as both a secreted cytokine and an intracellular nuclear factor. IL1β has a putative NLS (PKKKMEK) and it is known that mature (cleaved) IL1β can translocate to the nucleus with the receptor after IL1R1 mediated endocytosis. This translocation is believed to be receptor dependent as mutation of the NLS did not effect IL1β nuclear localization in this system.²⁶ It is very unlikely that interaction with IL1R1 is responsible for the localization of proIL1β to the nucleus since the proform of IL1β does not bind to IL1R1.²⁷ Recently, while investigating the nuclear localization of proIL1α in microglia, Lusheshi et al. detected nuclear proIL1β as well.²⁸ They found that proIL1a nuclear translocalization was NLS mediated and involves Ran-dependent active transport. In contrast, proIL1β nuclear translocalization was passive and NLS independent.²⁸

In order to specifically address the role of nuclear IL1β, we used genomic and bioinformatics approaches to begin to address the possibility that nuclear proIL1β in MCF10A-C/EBPβ2 cells was positioned in such a way as to contribute to the cancer phenotype by chromatin remodeling. ChIP-chip data indicates proIL1β is bound at distinct, non-random locations along the chromosome. ProIL1β is more tightly bound than histone H2A at the majority of these sites, but the details of how this cytokine binds to chromatin are presently still obscure. The IL-1 like cytokine IL33 contains a short (12 aa) chromatin-binding motif (CBM) that docks into an acidic pocket formed by the histone H2A-H2B dimer at the nucleosome surface.²² By interacting with nucleosomes, IL33 was shown to regulate chromatin compaction. However, the IL33 CBM, IL33_40–58, is not well conserved (4 out of 9 aa) in IL1β (or IL1α whose mechanism of chromatin interaction is also not well established). Thus, future experiments will be required to identify the IL1β CBM and to determine whether IL1β also interacts with core histones and/or some other chromatin-associated proteins.

As proIL1β may be altering chromatin architecture, via a yet to be elucidated mechanism, it is of interest that cancer associated pathways and certain genes known to be involved in metastasis initiation, including the processes of invasion, angiogenesis and EMT, were correlated with proIL1β sites (Table 2). In this setting, C/EBPβ2 upregulation of proIL1β may result in association of proIL1β with the chromatin and/or chromatin remodeling complexes facilitating changes in gene expression related to initiating EMT and promoting metastatic initiation, and to a lesser degree, metastatic progression and metastatic virulence. The location of proIL1β binding sites relative to genes involved in metastasis is intriguing, given that IL1β is a well established target gene for C/EBPβ in multiple cells types and C/EBPβ has recently emerged, along with the transcription factor, STAT3, as synergistic initiators and master regulators of mesenchymal transformation in malignant glioma.²⁹ However, the genomic-wide position analysis at best provides only circumstantial evidence. It is possible that IL1β knock-down would result in further insight into the role of IL1β in MCF10A-C/EBPβ2 cells. However, a key caveat is that IL1β knockdown would result in the loss of both nuclear and cytoplasmic IL1β, rendering it impossible to determine if the effects were due to nuclear IL1β specifically. Unfortunately, multiple attempts using both shRNA and siRNA technologies to knock-down IL1β have been unsuccessful to date although GAPDH knock-down was achieved. Given these limitations, elucidating the exact mechanism by which nuclear proIL1β influences chromatin conformation and contributes to the changes in EMT pathway genes seen in MCF10A-C/EBPβ2 cells will require significant future studies employing both biochemical and molecular mutagenesis approaches.

The accumulation of proIL1β in MCF10A-C/EBPβ2 cells raises the question of why the majority of this protein is not processed. Once IL1β is synthesized, inflammasome must be assembled and activated for proIL1β to be cleaved by caspase-1.³⁰ Multiple negative regulators of the inflammasome activation are now known; however, based on our genomic profiling data none of these negative regulators are expressed in MCF10A or MCF10A-C/EBPβ2 cells at a significant level (Suppl. file 3). Although caspase-1 is expressed, other known components of inflammasome complexes are not expressed at a significant level. These include: NALP1, NALP3, CARDINAL, caspase-5 and PYCARD (Suppl. file 3). The very low steady state mRNA level and/or lack of these components may explain the absence of the 17 kD protein in these cells.

It is notable that PYCARD, an important component of both the NALP1 and NALP3 inflammasomes, is silenced due to extensive promoter methylation in nearly half of breast cancer cell lines and over a third of human breast tumors.³¹ IL1β is present in over 90% of invasive breast carcinomas.³ It is, therefore, reasonable to expect that a subset of these breast cancers lack PYCARD expression due to promoter methylation, and therefore IL1β may be present in the proform. We demonstrate here that nuclear IL1β is present in some human breast cancer samples by immunohistochemical analysis. Further studies need to be done, especially to determine if lack of PYCARD is coincident with nuclear IL1β in any or all of these tumors. PYCARD promoter methylation has been observed in many other cancer types.^32–36 When combined with the known expression of IL1β in many of these same cancer types, it is possible proIL1β may be present in these cancers as well. These results establish proIL1β as an intracellular nuclear factor relevant for further study.

Methods

Cell culture and treatments.

Establishment and maintenance of MCF10A-C/EBPβ2 and MCF10A cells has been previously described.¹⁴ Recombinant human 17 kD IL1β (R&D Systems) or IL1R1 antagonist (IL1ra) (Kineret; Biovitrum) were added at indicated concentrations.

Genomic profiling.

Total RNA was isolated using the RNeasy Mini kit and RNase-Free DNase kit (Qiagen). RNA was submitted to the Vanderbilt Microarray Shared Resource for quality assurance and hybridization to Affymetrix GeneChip U133 PLUS 2.0 arrays microarray analysis. The resulting data has been submitted to Gene Expression Omnibus (GEO) under accession number GSE15065. GeneChip Operating System was used to grid images and generate .CEL and .CHP files for further analysis. Three independent replicates were performed. CEL files were imported in GeneSpring 7.0 (Agilent Technologies) and transformed by RMA (Robust Multichip Analysis). All probesets showing at least a 2 fold change in one of the C/EBPβ2 overexpressing MCF10A cells compared to parental MCF10A were tested with a Welch's t-test and a p value cutoff of 0.05.

Whole cell lysates, cell fractionation and immunoblot analysis.

Whole cell or nuclear extracts were prepared as preciously described.¹² Nuclear extraction in buffers of increasing ionic strength was performed in a similar manner, except that the sample was subjected to increasingly high ionic strength in a stepwise manner beginning at 0.1 M NaCl and ending at 1.0 M NaCl. After separation by electrophoresis the proteins were transferred to an Immobilon P filter and processed as previously described.¹³ The following primary antibodies were used anti-IL1β (MAB201 and AB-201-NA both from R&D Systems), anti-CREB (Santa Cruz), anti-H2A (Active Motif) and anti-β-tubulin (SIGMA).

Enzyme-linked immunosorbent assay (ELISA).

Fresh growth media was placed on 40–50% confluent cell cultures and conditioned for 24 hours. IL1β in conditioned media from cultured cells was determined by using the Quantikine HS IL-1b Immunoassay ELISA (R&D Systems) according the manufacturer's directions.

Indirect immunostaining and image acquisition.

Indirect immunostaining was performed as previously described with the following modifications.¹⁴ Anti-IL1β antibody (MAB201, R&D Systems) was used as the primary at a dilution of 1:500 and Hoechst solution (bisBenzimide) was used at 1 mg/ml to fluorescently label the nuclear compartment of the cells. The cells were visualized on a Leica DM IRB Inverted Microscope equipped with a Nikon DXM1200C camera.

Chromatin immunoprecipitation and promoter array.

MCF10A-C/EBPβ2 cells were fixed in 1% formaldehyde, snapfrozen and shipped to Genpathway where anti-IL1β (MAB201 R&D systems) was used to precipitate proIL1β bound chromatin. ProIL1β binding to human promoters was assessed using the Human Promoter 1.0 R Array (Affymetrix). The Cell Intensity Files were analyzed using Affymetrix' Tiling Analysis Software (TAS). TAS is used to generate signal values for all the probes on the arrays. Ratios are then generated by applying averaging and ranking steps.

Bioinformatic analysis of ChIP-chip data.

The following approach was used to determine whether the clustering of pro-IL1β in the genome was statistically significant. Given a cluster, we can calculate the range of the cluster, which is: Range = max(location) − min(location). Based on a chip with 35 bp probe spacing, we can set k = 35. Assuming there are m significant binding sites within this cluster, we can calculate the probability of m significant binding sites within a range as (where n = as.integer(Range/k):

$$prob=\left(\begin{array}{c}n\\ m\end{array}\right){\left(\frac{1}{n}\right)}^m$$

PANTHER pathway analysis was performed as previously described.³⁷ Next, to determine the statistical probability that proIL1β was associated with tumor initiation and various metastatic steps over-representation analysis (ORA) was performed. The ORA approach was previously published by Backes et al.³⁸ Based on the array description, there are approximately m = 4,600,000 probes in the array. From these probes, n = 204 significant binding sites were selected. For a given pathway, assuming there are l genes related to l binding sites, then there are k genes related to k binding sites for that pathway within the n = 204 significant binding sites. Therefore, we can calculate the p values for each of these pathways by the following formula:

$$P_C=\{\begin{array}{c}{\displaystyle \sum _{i=k}^{\text{K}}\frac{\left(\begin{array}{c}l\\ i\end{array}\right)\left(\begin{array}{c}m-l\\ n-i\end{array}\right)}{\left(\begin{array}{c}m\\ n\end{array}\right)}ifk\prime <k}\\ {\displaystyle \sum _{i=0}^k\frac{\left(\begin{array}{c}l\\ i\end{array}\right)\left(\begin{array}{c}m-l\\ n-i\end{array}\right)}{\left(\begin{array}{c}m\\ n\end{array}\right)}ifk\prime \ge k}\end{array}$$

Where K = min (n,l).

The R package R.basic was used to conduct the calculation.

Immunohistochemistry.

Breast tissue microarrays (Cybrdi, Inc.) were de-waxed, rehydrated, and subjected to thermal antigen retrieval in Retrievit Target Retrieval pH 4 according to the manufacturer's instructions using a microwave pressure cooker (InnoGenex). Sections were incubated with primary anti-IL1β antibody (AB-201-NA, R&D Systems), followed by incubation with biotinylated anti-goat antibodies (R&D Systems). Sections were incubated with avidin-peroxidase (Vector Laboratories), followed by DAB substrate (Invitrogen) and mounted. Using secondary antibody only resulted in no staining. The samples were imaged on an Olympus BX 51 microscope.

Abbreviations

C/EBPβ

CCAAT/enhancer binding protein-β

ChIP

chromatin immunoprecipitation

ELISA

enzyme-linked immunosorbent assay

EGF

epidermal growth factor

EMT

epithelial to mesenchymal transition

IL

interleukin