Coffee Polyphenols Change the Expression of STAT5B and ATF-2 Modifying Cyclin D1 Levels in Cancer Cells

Background. Epidemiological studies suggest that coffee consumption reduces the risk of cancer, but the molecular mechanisms of its chemopreventive effects remain unknown. Objective. To identify differentially expressed genes upon incubation of HT29 colon cancer cells with instant caffeinated coffee (ICC) or caffeic acid (CA) using whole-genome microarrays. Results. ICC incubation of HT29 cells caused the overexpression of 57 genes and the underexpression of 161, while CA incubation induced the overexpression of 12 genes and the underexpression of 32. Using Venn-Diagrams, we built a list of five overexpressed genes and twelve underexpressed genes in common between the two experimental conditions. This list was used to generate a biological association network in which STAT5B and ATF-2 appeared as highly interconnected nodes. STAT5B overexpression was confirmed at the mRNA and protein levels. For ATF-2, the changes in mRNA levels were confirmed for both ICC and CA, whereas the decrease in protein levels was only observed in CA-treated cells. The levels of cyclin D1, a target gene for both STAT5B and ATF-2, were downregulated by CA in colon cancer cells and by ICC and CA in breast cancer cells. Conclusions. Coffee polyphenols are able to affect cyclin D1 expression in cancer cells through the modulation of STAT5B and ATF-2.


Introduction
Polyphenols are the most abundant antioxidants in the diet. Their main dietary sources are fruits and plant-derived beverages such as fruit juices, tea, coffee, and red wine. Current evidence strongly supports a contribution of polyphenols to the prevention of cardiovascular diseases, cancers, and osteoporosis suggesting a role of these antioxidants in the prevention of neurodegenerative diseases and diabetes mellitus [1].
It is well established that polyphenol ingestion results in an increase of the plasma-antioxidant capacity. However, there is still some uncertainties about their efficiency to enhance the protection of cellular components, such as lipids or DNA, against oxidative stress in humans [2]. Polyphenols and other antioxidants were thought to protect cell constituents against oxidative damage by scavenging free radicals. However, this concept now appears to be an oversimplified view of their mode of action [3]. More likely, cells respond to polyphenols mainly through direct interactions with receptors or enzymes involved in signal transduction, which may result in modification of the redox status of the cell and may trigger a series of redox-dependent reactions [4]. This could also apply to the anticarcinogenic effects of polyphenols, which properties may be explained by many different mechanisms.
Hydroxycinnamic acids are a major class of polyphenols found in almost every plant [2]. The major representative of hydroxycinnamic acids is caffeic acid, which occurs in food mainly as an ester with quinic acid named chlorogenic acid (5-caffeoylquinic acid). Coffee is a major source of chlorogenic acid in the human diet; the daily intake in coffee drinkers is 0.5-1 g whereas coffee abstainers will usually ingest <100 mg/day. Studies have shown that approximately the 33% of ingested chlorogenic acid and the 95% of caffeic acid are absorbed intestinally [5]. Thus, about two-thirds of ingested chlorogenic acid reach the colon where it is probably metabolized to caffeic acid [6].
Bioavailability data suggest that the biological effects of chlorogenic acid would become apparent after its metabolism to caffeic acid, and hence the need of studying the effects of this acid. Chlorogenic acid and caffeic acid are 2 Oxidative Medicine and Cellular Longevity antioxidants in vitro [7], and they might inhibit the formation of mutagenic and carcinogenic N-nitroso compounds since they are inhibitors of the N-nitrosation reaction in vivo [8]. Furthermore, chlorogenic acid can inhibit DNA damage in vitro [9] as it inhibits lipid peroxidation-induced DNA adduct formation [10] and suppresses reactive oxygen species-mediated nuclear factor (NF-κB), activator protein-1 (AP-1), and mitogen-activated protein kinase activation by upregulating antioxidant enzymes [11]. These studies suggested that coffee polyphenols are potent chemopreventive agents.
Recent meta-analyses demonstrate inverse associations between coffee intake and the risk of colon, liver, breast, and endometrial cancer [12][13][14][15]. Moreover, in prospective population-based cohort studies, the inverse association between coffee consumption and risk of cancer has been shown. The group of Naganuma [16] found that the consumption of at least one cup of coffee per day was associated with a 49% lower risk of upper gastrointestinal cancer in a Japanese population, while Wilson and collaborators [17] found that men who regularly drink coffee appeared to have a lower risk of developing a lethal form of prostate cancer. The lower risk was evident when consuming either regular or decaffeinated coffee. It has been proposed that the inverse association between coffee intake and colon cancer could be explained, at least in part, by the presence of chlorogenic acid in coffee [18]. Ganmaa et al. [19] observed a general protective effect of caffeine intake on breast cancer risk for both ER subtypes, but the effect was only found to be significant for ER-positive breast cancers. In this study, the association between caffeine and breast cancer was stronger among postmenopausal women with estrogen-receptor and progesterone-receptor-positive breast cancer than those with estrogen-receptor and progesterone-receptor negative breast cancer [19]. In another study, coffee drinking specifically reduced the risk of developing ER-negative breast cancer but not ER-positive breast cancer [20].
Although there is enough evidence from epidemiological data supporting that coffee seems to reduce the risk of certain cancers, the molecular mechanisms underlying the chemopreventive effects of coffee remain unknown. For this reason, the aim of our study was to determine the effect at the molecular level of coffee polyphenols at low concentrations equivalent to one cup of coffee, using as a model a human colon cancer cell line HT29 in a nutrigenomic approach. Furthermore, the effect of coffee polyphenols was also evaluated in breast cancer cells.

Microarray Data Analyses.
Quantification was carried out with GeneSpring GX v.11.5.1 software (Agilent Technologies), which allows multifilter comparisons using data from different experiments to perform the normalization, generation of lists, and the functional classification of the differentially expressed genes. The input data was subjected to preprocess baseline transformation using the Robust Multiarray Average summarization algorithm using the median of control samples. After grouping the triplicate of each experimental condition, list of differentially expressed genes could be generated by using volcano plot analysis. The expression of each gene is reported as the ratio of the value obtained after each condition relative to control condition after normalization and statistical analysis of the data. The corrected P value cutoff applied was of <0.05; then the output of this statistical analysis was filtered by fold expression, selecting specifically those genes that had a differential expression of at least 1.3-fold. Gene classification was established by the Gene Ontology database.

Common
Genes between ICC and CA Treatments. Common genes were selected from the lists of differentially expressed genes for each treatment using Venn-Diagrams. The newly generated list contained both over and underexpressed genes.

Generation of Biological Association Networks.
BANs were constructed with the aid of the Pathway Analysis within the GeneSpring v.11.5.1 (Agilent) as described in Selga et al. [22] with the list of common genes differentially expressed in both treatments. A filtered screening was processed by the program between our data and bibliographic interaction databases up to a total of 100 related genes. Network associations were confirmed in the literature.

RT Real-Time PCR.
Total RNA was extracted from HT29 cells using Ultraspec (Biotex) in accordance with the manufacturer's instructions.
Complementary DNA was synthesized as described in Selga et al. [23] and the cDNA product was used for amplification by real time PCR. STAT5B and ATF-2 mRNA levels were determined in an ABI Prism 7000 Sequence Detection System (Applied Biosystems) using 3 μL of the cDNA reaction and the assays-on-demand Hs00560035 m1 for STAT5B, Hs00153179 ml for ATF-2, and Hs00356991 m1 for APRT (all from Applied Biosystems). APRT mRNA was used as an endogenous control. The reaction was performed following the manufacturers recommendations. Fold changes in gene expression were calculated using the standard ΔΔCt method.

Western
Blot. Whole extracts were obtained from 2.5 × 10 6 control or treated cells according to Selga et al. [23]. Five μL of the extract was used to determine protein concentration by the Bradford assay (Bio-Rad). The extracts were frozen in liquid N 2 and stored at −80 • C. Total extracts (50 μg) were resolved on SDS-polyacrylamide gels and transferred to PVDF membranes (Immobilon P, Millipore) using a semidry electroblotter.
The SNAP i.d. protein detection system technology (Millipore) was used to probe the membranes. This system applies vacuum through the membrane to actively drive reagents to protein locations, unlike the traditional technique of diffusion over the membrane as a reagent transport. Table 1 compiles the antibodies used in the different determinations.

Statistical Methods.
For the RT-PCR and Western blot analyses, values are expressed as the mean ± SE of three different experiments. Data were evaluated by unpaired Student's t test, and analyses were performed using the PASW Statistics v. 18.0.0. software.

Effect of ICC and CA Incubations in HT29 Gene
Expression. The expression profile of over 47,000 transcripts and variants included in the microarray HG U133 plus 2.0 from Affymetrix was compared between HT29 control cells and cells incubated with either CA or ICC, at nontoxic concentrations for 24 h. GeneSpring GX software v.11.5.1 was used to analyze the results. A list of differentially expressed genes by 1.3-fold with a P value cutoff of <0.05 was generated as described in Methods. When HT29 cells were incubated with ICC, 57 genes were overexpressed whereas 161 genes were underexpressed. Among the overexpressed genes, 24% belonged to the Transcription factors category and 19% to Cell cycle or to Biosynthetic processes. Within the underexpressed genes, the category corresponding to cell cycle was the most affected (53% of the genes) followed by Transcription factors (19%) and Biosynthetic processes (12%). Upon incubation with CA, 12 genes were overexpressed whereas 32 genes were underexpressed. Among the overexpressed genes, 33% belonged to the Transcription factors category, 25% to Cell cycle, and 16,7% to Biosynthetic processes or immune response. Within the underexpressed genes, again the category corresponding to Cell cycle was the most affected (30% of the genes) followed by Biosynthetic processes (15%) and Transcription factors (12%). The lists of differentially expressed genes are presented as Tables 2, 3, 4, and 5. The data presented in this work have been deposited in the Gene Expression Omnibus (GEO, http://www.ncbi.nlm.nih.gov/geo/) and are accessible through GEO series accession number [GSM867162].

Generation of Biological Association Networks.
A Biological Association Network (BAN) was constructed using the Pathway Analysis within GeneSpring v.11.5.1 as described in Methods using as the starting list the common genes differentially expressed upon incubation with CA and ICC. This list included five overexpressed genes and twelve underexpressed genes ( Table 6). In the generated network, signal transducer and activator of transcription 5B (STAT5B) and activating transcription factor 2 (ATF-2) appeared as highly interconnected nodes ( Figure 1). These two main nodes were selected for further validations. STAT5B was overexpressed with respect to the control by 23.8% in cells treated with ICC and by 33.4% in cells treated with CA, whereas ATF-2 was found underexpressed in HT29 incubated with ICC (32.5% decrease compared to the control) and with CA (26% decrease).

Validation of STAT5B and ATF-2 Changes at the mRNA and Protein
Levels. STAT5B overexpression in HT29 cells upon incubation with CA and ICC was confirmed at the mRNA (1.16-and 1.3-fold compared to the control, respectively) and protein levels (1.5-and 1.2-fold compared to the control, respectively) (Figures 2(a) and 2(c)). In the case of ATF-2, the changes in mRNA levels were confirmed for both CA and ICC (0.88-and 0.86-fold compared to the control, respectively), whereas the decrease in protein levels was only observed in CA-treated cells (0.62-fold compared to the control) (Figures 2(b) and 2(d)).

Expression of Cyclin D1 upon Incubation with ICC and
CA. Cyclin D1 is overexpressed at the mRNA and protein level in over 50% of the breast cancers either in the presence or absence of gene amplification, and it is one of the most commonly overexpressed proteins in breast cancer [24,25]. Cyclin D1 transcription is regulated by STAT5 [26][27][28][29] and ATF-2 [30][31][32]. : Biological association network (BAN) of differentially expressed genes in common between CA and ICC. The list of common genes between both treatments was used to construct a BAN with the Pathway Analysis software within GeneSpring v.11.5.1. An expanded network was constructed by setting an advanced filter that included the categories of binding, expression, metabolism, promoter binding, protein modification, and regulation. Only proteins are represented. The BAN shows the node genes STAT5B and ATF-2 that were further studied. We analyzed the levels of cyclin D1 by western blot in MCF-7 and HT29 cells upon incubation with ICC and CA. As shown in Figure 3(a), incubation of MCF-7 cells with either CA and ICC led to a drastic decrease in the levels of cyclin D1 protein, together with an increase in the levels of STAT5B, but not to a decrease in the levels of ATF-2. In HT29 cells, incubation with CA did not affect cyclin D1 levels, whereas the presence of ICC led to an increase in cyclin D1 levels 3 (b).

Discussion
In this work we analyzed the gene expression profile of human cancer cells treated with either ICC or CA. Caffeic acid was chosen since it is the main representative of hydroxycinnamic acids. Using microarrays we identified the differential expression of specific genes involved in several biological pathways. The changes in mRNA expression of two outlier genes, STAT5B and ATF-2, observed in the microarrays were confirmed by RT real-time PCR, and the changes in protein levels were also analyzed by Western blot. The selection of STAT5B and ATF-2 was made according to the results obtained in the construction of a biological association network. Finally, the modulation of cyclin D1, a target of STAT5B and ATF-2 transcription factors, upon incubation with coffee polyphenols was also established.
We show that ICC and the amount of CA of one cup of coffee are able to induce STAT5B mRNA and protein Oxidative Medicine and Cellular Longevity     Oxidative Medicine and Cellular Longevity    The table shows the list of underexpressed genes by 1.3-fold with a P value < 0.05 obtained in cells treated with instant caffeinated coffee and includes the gene symbol for all genes, and their associated description. The ratio columns correspond to the absolute fold change in expression relative to the control group and the type of regulation (down: downregulation). The table shows the list of overexpressed genes by 1.3-fold with a P value < 0.05 obtained in cells treated with caffeic acid and includes the gene symbol for all genes, their associated description. The ratio columns correspond to the absolute fold change in expression relative to the control group and the type of regulation (up: upregulation). levels in HT29 cells. STAT5 was originally described as a prolactin-induced mammary gland factor [33]. The cloning of two closely related STAT5 cDNAs, from both mouse and human cDNA libraries, showed two distinct genes, STAT5A and STAT5B that encoded two STAT5 proteins [34][35][36][37].
In addition to prolactin, STAT5 proteins are activated by a wide variety of cytokines and growth factors, including IL-2, IL-3, IL-5, IL-7, IL-9, IL-15, granulocyte-macrophage colony-stimulating factor, erythropoietin, growth hormone, thrombopoietin, epidermal growth factor, and plateletderived growth factor. The key function of STAT5B is to mediate the effects of growth hormone [38,39]. Modulation of STAT5 levels or transcriptional activity has already been described in cells treated with natural compounds such as nobiletin, a citrus flavonoid [40], thea flavins [41], and silibinin, a natural polyphenolic flavonoid which is a major bioactive component of silymarin isolated from Silybum marianum [42]. Furthermore, it has been reported that butein, the major biologically active polyphenolic component of the stems of Rhus verniciflua, downregulated the expression of STAT3-regulated gene products such as Bcl-xL, Bcl-2, cyclin D1, and Mcl-1 [43]. STAT5B participates in diverse biological processes, such as growth development, immunoregulation, apoptosis, reproduction, prolactin pathway, and lipid metabolism. STAT5B deficiency is a recently identified disease entity that involves both severe growth hormone-resistant growth failure and severe immunodeficiency [44][45][46]. The induction of STAT5B expression upon incubation with CA and ICC could represent a nutritional tool to upregulate this transcription factor and suggests novel research strategies for natural therapies in Crohn's disease and inflammatory bowel disease in which STAT5B appears to maintain the mucosal barrier integrity and tolerance [47,48]. In colorectal cancer both STAT5a and STAT5b play important roles in progression and downregulation of both STAT5A and STAT5B results in a gradual decrease in cell viability, predominantly attributed to G1 cell cycle arrest, and apoptotic cell death [49]. In this context the increase in STAT5B caused by ICC and CA would have a negative effect on colorectal cancer patients, as it would trigger cell proliferation and survival.
In human breast cancer, STAT5A/B has been shown a dual role in the mammary gland as an initiator of tumor formation as well as a promoter of differentiation of established tumors. STAT3, STAT5A, and STAT5B are overexpressed or constitutively activated in breast cancer [50][51][52] and active STAT5A/B in human breast cancer predicted favorable clinical outcome [53]. Prolactin receptor signal transduction through the Jak2-STAT5 pathway has been considered to be essential for proliferation and differentiation of normal mammary epithelial cells [54][55][56]. It has been shown that the levels of NUC-pYSTAT5 decreased as breast cancer progressed from normal to in situ, to invasive, and then to nodal metastases [57]. Additionally Peck et al. [57] found that the absence of detectable NUC-pYStat5 in tumors of patients how where under antiestrogen therapy was associated with poor breast cancer-specific survival. We analyzed STAT5B modulation through the PRL pathway in response to coffee polyphenols in a breast cancer cell line.
The MCF-7 cell line was chosen because expression of the prolactin receptor is more often found in estrogen receptorpositive breast tumors [58]. In our conditions, incubation with CA and ICC led to an increase in STAT5B protein levels in MCF-7 cells, and this result could be the basis for a possible inclusion of coffee polyphenols in the diet of breast cancer patients. ATF-2 is a member of the ATF-cAMP response elementbinding protein (CREB) family of transcription factors that can bind to the cAMP response element (CRE) found in many mammalian gene promoters [59,60]. ATF-2 exhibits both oncogenic and tumor suppressor functions [61]. CREs are found in several genes involved in the control of the cell cycle, for example, the cyclin D1 gene, and ATF-2 binding to this sequence stimulates the transcription of cyclin D1 [30,31]. ATF-2 mediated cyclin D1 promoter induction can be stimulated by a number of growthpromoting agents, such as estrogen [31], hepatocyte growth factor [62], and regenerating gene product [63]. ATF-2 has been correlated with proliferation, invasion, migration, and resistance to DNA-damaging agents in breast cancer cell lines.
The downregulation of ATF-2 expression after CA and ICC incubation in HT29 cells reported here is in accordance with the observed decrease in activity of ATF-2 in gastric cells when incubating with chlorogenic acid, the precursor of caffeic acid [64]. Surprisingly, the validation of the protein levels showed the upregulation of ATF-2 protein with ICC, but not with CA, both in HT29 and MCF-7 cells. This differential behavior could be due to other ICC components besides CA. In this direction Rubach et al. [64] reported a different response in ATF-2 activity after incubation of a gastric cell line with different coffee compounds. The presence of pyrogallol, catechol, βN-alkanoylhydroxytryptamides, and N-methylpyridinium increased ATF-2 activity, whereas chlorogenic acid and caffeine decrease it [64]. In our conditions incubation of HT29 cells with ICC caused a modest decrease in ATF-2 mRNA levels. However this effect was not translated at the protein level. We hypothesize that ICC contains other polyphenols in addition to caffeic acid that are able to increase ATF-2 protein levels through an increase of the translation of its mRNA, the increase of stability of the protein or an inhibition of its degradation. In this direction several plant polyphenols such as (-)-epigallocatechins-3-gallate (EGCG), genistein, luteolin, apigenin, chrysin, quercetin, curcumin, and tannic acid have been described to possess proteasomeinhibitory activity [65,66].
The regulation of ATF-2 transcriptional activity, mostly at the level of its phosphorylation status, has been described upon treatment of cancer cells with several natural compounds. In MCF-7 cells, the anticancer agent 3,30-Diindolylmethane, derived from Brassica vegetables, activates both JNK and p38 pathways, resulting in c-Jun and ATF-2 phosphorylation, and the increase of binding of the c-Jun-ATF-2 homodimers and heterodimers to the proximal regulatory element of IFN-γ promoter [67]. Biochanin-A, an isoflavone, existing in red clover, cabbage and alfalfa, has an inhibitory and apoptogenic effect on certain cancer cells by blocking the phosphorylation of p38 MAPK and ATF-2 in a dose-dependent fashion [68]. The JNK stress-activated pathway is one of the major intracellular signal transduction cascades involved in intestinal inflammation [69,70], and upregulation of ATF-2 has been shown in Crohn's disease [71,72]. Thus CA could represent potential therapeutical properties in different states of intestinal inflammation due to its combined effects on STAT5B and ATF-2 in HT29 cells.
Finally, the modulation of cyclin D1, a target of STAT5B and ATF-2 transcription factors, upon incubation with coffee polyphenols was established in colon and breast cancer cells. Cyclin D1 overexpression is common in colorectal cancer, but the findings regarding its prognostic value are conflicting. In a recent study, positive expression of cyclin D1 protein was detected in 95 of 169 colonic adenocarcinoma specimens, and increased cyclin D1 levels were associated with poorer prognosis [73]. Furthermore, there was a significant correlation between the positive expression of p-Stat5 and cyclin D1 in patients with colonic adenocarcinoma. However, in a second study, cyclin D1 overexpression was associated with improved outcome in a total of 386 patients who underwent surgical resection for colon cancer, classified as TNM stage II or III. Belt et al. [74] showed that low p21, high p53, low cyclin D1, and high AURKA were associated with disease recurrence in stage II and III colon cancer patients. In this context the effect of ICC on cyclin D1 levels could represent either a positive or a negative effect in colon cancer cells, depending on tumor progression. The increase in cyclin D1 levels could represent a marker of better outcome since it has been recently established that cyclin D1 expression is strongly associated with prolonged survival in male colorectal cancer and that lack of cyclin D1 is associated with a more aggressive phenotype in male patients [75]. However, several natural compounds such as anthocyanins, anthocyanidins, apigenin, luteolin, and fisetin have all been described to induce experimentally cell-cycle arrest and apoptosis through the decrease of cyclin D1 levels in HT29 cells [76][77][78][79][80]. In accordance to these data, the increase observed in cyclin D1 levels in HT29 cells upon incubation with ICC could probably be the consequence of the presence of different compounds other than polyphenols in ICC.
In MCF-7 breast cancer cells, cyclin D1 was downregulated upon incubation with coffee polyphenols. The rationale for the choice of MCF-7 cell line was based on the observation that although cyclin D1 overexpression is present across multiple histologic subtypes of breast cancer, it has been shown that the large majority of cyclin D1overexpressing breast cancers are ER positive [24,25,81]. Cyclin D1 overexpression has been reported between 40 and 90% of cases of invasive breast cancer, while gene amplification is seen in about 5-20% of tumors [24,[81][82][83]. In cyclin D1-driven cancers, blocking cyclin D1 expression by targeting the cyclin D1 gene, RNA, or protein should increase the chances for therapeutic success. Cell culture studies have raised the possibility that certain compounds might act in this way [84,85] and approaches to blocking cyclin D1 expression using antisense, siRNA, or related molecules specifically target the driving molecular lesion itself [86][87][88]. It is believed that compounds that modulate cyclin D1 expression could have a role in the prevention and treatment of human neoplasias. For instance, flavopiridol, a synthetic flavonoid based on an extract from an Indian plant for the potential treatment of cancer, induces a rapid decline in cyclin D1 steady-state protein levels [89]. Taking all these results together, inhibition of cyclin D1 expression appears to be a good approach for cancer treatment. In this direction our observation that coffee and caffeic acid are able to drastically reduce the expression of cyclin D1 in breast cancer cells could suggest that some coffee components could be used as a coadjuvant therapeutic tool in the treatment of breast cancer.