Mullerian Inhibiting Substance Promotes Interferon γ-induced Gene Expression and Apoptosis in Breast Cancer Cells*

This report demonstrates that in addition to interferons and cytokines, members of the TGFβ superfamily such as Mullerian inhibiting substance (MIS) and activin A also regulate IRF-1 expression. MIS induced IRF-1 expression in the mammary glands of mice in vivo and in breast cancer cells in vitro and stimulation of IRF-1 by MIS was dependent on activation of the NFκB pathway. In the rat mammary gland, IRF-1 expression gradually decreased during pregnancy and lactation but increased at involution. In breast cancer, the IRF-1 protein was absent in 13% of tumors tested compared with matched normal glands. Consistent with its growth suppressive activity, expression of IRF-1 in breast cancer cells induced apoptosis. Treatment of breast cancer cells with MIS and interferon γ (IFN-γ) co-stimulated IRF-1 and CEACAM1 expression and synergistic induction of CEACAM1 by a combination of MIS and IFN-γ was impaired by antisense IRF-1 expression. Furthermore, a combination of IFN-γ and MIS inhibited the growth of breast cancer cells to a greater extent than either one alone. Both reagents alone significantly decreased the fraction of cells in the S-phase of the cell cycle, an effect not enhanced when they were used in combination. However, MIS promoted IFN-γ-induced apoptosis demonstrating a functional interaction between these two classes of signaling molecules in regulation of breast cancer cell growth.


This report demonstrates that in addition to interferons and cytokines, members of the TGF␤ superfamily such as Mullerian inhibiting substance (MIS) and activin A also regulate IRF-1 expression. MIS induced IRF-1 expression in the mammary glands of mice in vivo and in breast cancer cells in vitro and stimulation of IRF-1 by MIS was dependent on activation of the NFB pathway. In the rat mammary gland, IRF-1 expression gradually decreased during pregnancy and lactation but increased at involution. In breast cancer, the IRF-1 protein was absent in 13% of tumors tested compared with matched normal glands. Consistent with its growth suppressive activity, expression of IRF-1 in breast cancer cells induced apoptosis. Treatment of breast cancer cells with MIS and interferon ␥ (IFN-␥) co-stimulated IRF-1 and CEACAM1 expression and synergistic induction of CEACAM1 by a combination of MIS and IFN-␥ was impaired by antisense IRF-1 expression. Furthermore, a combination of IFN-␥ and MIS inhibited the growth of breast cancer cells to a greater extent than
either one alone. Both reagents alone significantly decreased the fraction of cells in the S-phase of the cell cycle, an effect not enhanced when they were used in combination. However, MIS promoted IFN-␥-induced apoptosis demonstrating a functional interaction between these two classes of signaling molecules in regulation of breast cancer cell growth.
Mullerian Inhibiting Substance (MIS) 1 is a member of the TGF␤ family, a class of molecules that regulates growth, differentiation, and apoptosis in many cell types. In the male embryo, MIS causes regression of the Mullerian duct, the an-lagen of the Fallopian tubes, uterus, and the upper vagina (1). However, a postnatal role for MIS in males and females has yet to be clearly defined. MIS receptor mRNA in the mammary gland significantly diminishes during puberty when the ductal system branches and invades the adipose stroma and during the expansive growth at pregnancy and lactation, but is upregulated during involution, a time of regression and apoptosis (2,3). The inverse correlation between MIS type II receptor expression and various stages of mammary growth suggests that MIS-mediated signaling may exert an inhibitory effect on mammary gland growth. Consistent with this concept, MIS inhibited the growth of both estrogen receptor (ER)-positive and -negative breast cancer cells by inducing cell cycle arrest and apoptosis (4).
Type I (IFN-␣ and IFN-␤) and type II (IFN-␥) interferons are a family of antiviral cytokines that exhibit immunomodulatory and anti-proliferative effects (5). The antitumor effects of cytokines such as interleukin-12, in murine mammary carcinogenesis models correlate with high levels of serum IFN-␥ (6 -12). IFN-␥ induced tumor regression results from immune surveillance of tumor cells and from direct cytotoxic effects (13)(14)(15)(16)(17), which are evident from its ability to inhibit the growth of several tumor-derived cell lines (18,19) including breast cancer cells (20 -22). Intralesional injections of IFN-␣ and IFN-␥ into breast cancer patients with skin recurrences resulted in either complete or partial regression of the skin lesions but was associated with clinical toxicity in all patients (23). Thus identification of molecules that enhance the antitumor effects of IFN-␥ may render it effective at lower doses, reduce clinical toxicity associated with high concentrations of the drug, and expand their therapeutic applications.
IFN-␥-induced growth inhibition requires coordinate expression of specific genes. Interferon regulatory factor-1 (IRF-1) is robustly induced by both type I and type II interferons. In addition to its important role in innate and adaptive immunity (24), IRF-1 also plays a role in regulating the growth of different mammalian cell lines (25). Different aspects of the tumor suppressor function of IRF-1 may be explained, at least in part, by the observation that it induces several growth regulatory genes including those with anti-proliferative activity such as IFN␣/␤, p21, and the cell adhesion molecule CEACAM1 (carcinoembryonic antigen-related cell adhesion molecule) (25).
Using DNA microarrays to profile gene expression, we identified that treatment of breast cancer cells with MIS strongly induces the expression of IRF-1. Since interferons also strongly induce IRF-1 (15), we tested whether intersection of MIS and IFN-␥ signaling pathways would result in enhanced expression of downstream target genes and increased antiproliferative activity against breast cancer cells. In this report, we demonstrate that members of the TGF␤ superfamily including MIS induce IRF-1 expression in immortalized human mammary epithelial cells with characteristics of normal cells. Treatment of breast cancer cells with MIS and IFN-␥ led to synergistic induction of CEACAM-1 through an IRF-1-dependent mechanism. Furthermore, a combination of MIS and IFN-␥ led to a greater degree of growth inhibition compared with either agent alone due to enhanced apoptosis rather than a combinatorial effect on cell cycle progression.

EXPERIMENTAL PROCEDURES
Cell Culture and MTT Assays-Human breast cancer cell lines T47D and MDA-MB-468 were grown in Dulbecco's modified medium supplemented with 10% female fetal bovine serum, glutamine, and penicillin/ streptomycin. The human mammary epithelial cell line MCF10A was grown in mammary epithelial growth medium (MEGM, Clonetics) supplemented with 100 ng/ml of cholera toxin (Calbiochem). Human recombinant MIS (26) was collected from growth media of Chinese hamster ovary cells transfected with the human MIS gene and purified as described (26). Recombinant human IFN-␥ was purchased from Sigma and IFN-␤ and activin A were from R&D systems, Inc. TGF␤ was a kind gift from Dr. Anita Roberts.
Estimation of cell growth was based on the colorimetric reduction of a yellow tetrazolium salt, MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide), to a purple formazan by viable cells. The MDA-MB-468 cell suspension (3000 cells/well) was transferred to a 96-well microtiter plate. MIS, IFN-␥, or both were added to the wells once on day 0 at concentrations indicated in the figure legends. The number of viable cells was estimated by adding 30 l of MTT solution (5 mg/ml in phosphate-buffered saline). Following 3 h of incubation at 37°C, the stain was eluted into 200 l of Me 2 SO. The optical densities were quantified at a test wavelength of 550 nm and a reference wavelength of 630 nm on a multiwell spectrophotometer. Statistical analysis was done using Student's t test.
T47D cells expressing antisense IRF-1 mRNA were generated by co-transfecting 0.5 g of hygromycin resistance plasmid and 10 g of pCDNA 3.1 expressing the IRF-1 transcript in the antisense orientation. Cells were maintained in medium containing 100 g/ml of hygromycin (Roche Applied Science), and clones expressing antisense IRF-1 were identified by Northern blot analysis.
Western Blot Analysis-Expression of protein in cells was analyzed by Western blot using the rabbit anti-IRF-1 (Santa Cruz Biotechnology), rabbit antiphospho-STAT1, and rabbit anti-STAT1 (Cell Signaling) antibodies according to the protocol described (27). Nuclear and cytoplasmic protein fractions from T47D cells were isolated according to the protocol described below.
NFB and STAT Electrophoretic Mobility Shift Assays-T47D cells were grown to 70% confluence and treated with indicated concentrations of rhMIS or IFN-␥. Cells were harvested in cold PBS, resuspended in 1 ml TKM 10:10:1 (10 mM Tris, pH 8.0, 10 mM KCl, and 1 mM MgCl 2 ) and lysed with 0.1% Triton X-100. The cytoplasmic fraction was stored frozen for Western blot analysis of STAT proteins. Nuclei were pelleted by centrifugation at 5,000 rpm at 4°C, and proteins were extracted in buffer containing 10 mM HEPES pH 7.0, 350 mM NaCl, and 1 mM EDTA. 3 g of protein were used in 25-l binding reactions containing 10 mM HEPES pH 7.0, 70 mM NaCl, 0.1% Triton X-100, and 4% glycerol. NFB (Promega), and SIE (Stat Inducing Element; Geneka) oligonucleotides were 5Ј-end-labeled with 32 P and DNA protein complexes were resolved on 4% native polyacrylamide gels. Supershift experiments were performed by adding 1 g of rabbit anti-p65 or p50 antibodies (Santa Cruz Biotechnology) or rabbit anti-STAT1, STAT3, or STAT5␣ (Santa Cruz Biotechnology) antibodies to the binding reactions.
Animals and MIS Treatment-All animals were cared for and experiments were performed under AAALAS approved guidelines using protocols approved by the Institutional Review Board-Institutional Animal Care and Use Committee of the Massachusetts General Hospital. Ketamine/xylazine (100/10 mg/kg) was used for anesthesia. To study the effect of rhMIS on the mammary gland, adult female C3H mice (8-weekold; average weight 25 grams) were obtained from the Edwin L. Steele Laboratory, Massachusetts General Hospital, Boston, MA. Each animal was injected intraperitoneally with 100 g of rhMIS or phosphatebuffered saline (vehicle control). Breast tissue was harvested bilaterally from each animal for RNA isolation. Blood was drawn from the animals at the time of tissue harvest to determine the circulating level of rhMIS using MIS-ELISA.
IRF-1 expression analysis in the rat breast during perinatal morphogenesis was done using Sprague-Dawley rats. To analyze expression during lactation and involution, after the pups were born (postdelivery), some animals were housed with the pups (lactating) while others were weaned 2 days after lactation (weaned).
RNA Analysis-Total RNA from T47 cells treated with MIS for 0 and 1 h was isolated using RNA STAT-60 and sent to the Hartwell Center for Bioinformatics and Biotechnology at St. Jude Children's Research Hospital for profiling gene expression using HG-U95Av2 oligonucleotide arrays (Affymetrix) containing ϳ12,500 full-length annotated genes together with additional probe sets designed to represent EST sequences. The EST clones to detect the expression of IRF-1 and CEACAM1 were purchased from Incyte Genomics Inc. For Northern blot analysis, equal amounts of RNA were separated on a formaldehyde gel, transferred to Hybond-N membrane (Amersham Biosciences) and probed with either IRF-1 or CEACAM1.
Cell Cycle Analysis and Apoptosis Assays-The cell cycle distribution of untreated and MIS and IFN-␥-treated cultures was analyzed by fluorescence activated cell sorting (FACS). Cells were detached with PBS/EDTA, fixed in 95% ethanol and treated with propidium iodide and RNase A. Flow cytometric analysis was performed on a Becton Dickinson FACScan flow cytometer.
To measure the apoptosis, cells treated with MIS and IFN-␥ were immunostained with a FITC-conjugated anti-annexin V antibody, counterstained with DAPI and analyzed by FACS. The effect of IRF-1 on apoptosis, was analyzed by transiently transfecting cells in logarithmic growth phase with either the empty vector (4 g) or IRF-1 (4 g), along with a plasmid encoding the cell surface marker CD20 (1 g) as described in (28). Cells were harvested 72 h after transfection and stained for CD20 and annexin V. Apoptosis in CD20-positive cells was determined by FACS analysis.
Immunohistochemical Analysis of IRF-1-Twenty-three cases of human breast carcinoma with adjacent uninvolved breast parenchyma were selected from the files of the Massachusetts General Hospital (MGH) Pathology Department according to protocols approved by the Human Research Committee at MGH. These included 13 poorly differentiated (histologic grade 3/3), 5 moderately differentiated (grade 2/3) and 3 well-differentiated (grade 1/3) ductal adenocarcinomas. Patient age ranged from 34 to 81 years at the time of the procedure. Immunohistochemistry was performed on formalin-fixed, paraffin-embedded tissue sections according to standard procedures. Deparaffinized tissue sections were treated with 3% hydrogen peroxide for 30 min to inhibit endogenous peroxidase. After the sections had been microwaved in a 10 mM citrate buffer solution at 100 watts for 25 min, they were incubated with 2% normal goat serum for 30 min to block any nonspecific protein binding. The sections were incubated with rabbit antibodies against IRF-1 (Santa Cruz Biotechnology) at 4°C overnight at a 1:400 dilution. After incubation with biotinylated antibody and peroxidase-labeled streptavidin, the staining was developed by reaction with 3. 3Ј-diamino benzidine for 5 min (ABC kit, Vector Laboratories, Burlingame, CA), and used according to the manufacturer's instructions. The sections were lightly counterstained with hematoxylin. Negative staining of tumor was determined only in the presence of stained uninvolved glands on the same sections. Tumors were considered IRF-1-negative if no signal was detectable in Ͼ95% of the tumor. The specificity of IRF-1 induction by MIS was tested by treating cells with heat inactivated MIS or with the affinity-purified noncleavable, biologically inactive form of rhMIS (L9, 35 nM) that does not induce the regression of the Mullerian duct in organ culture assays (29) or inhibit the growth of T47D cells (4) (Fig. 1A, lower right panel). MIS-mediated induction of IRF-1 mRNA was also observed in MCF10A cells (Fig. 1B, left panel), a non-tumorigenic human mammary epithelial cell line with normal karyotype derived from a patient with fibrocystic breast disease (30). Furthermore, both activin and TGF␤ induced IRF-1 in MCF10A cells (Fig. 1B, right panels). Thus IRF-1 expression in mammary epithelial cells may be under the regulation of multiple members of the TGF␤ family including MIS.

Members of the TGF␤ Family Induce IRF-1 Expression-
We next determined whether exposure of mammary glands to exogenous rhMIS would result in the induction of IRF-1 in vivo. Intraperitoneal injection of rhMIS into mice induced IRF-1 expression in the mammary glands compared with PBSinjected controls (Fig. 1C). The serum rhMIS levels averaged 2-4 g/ml in the animals as measured by ELISA (26).
Expression of IRF-1 in the Rat Mammary Gland and in Human Breast Cancer-Expression analysis of IRF1 in the mammary glands of virgin, pregnant, lactating, and weaned rats demonstrated that IRF-1 mRNA was detectable in the virgin animals but gradually declined during pregnancy (G5-G21) and reached a nadir at late pregnancy (G17-G21) and lactation (PD0-PD10: lactating). Expression rebounded in the mammary glands of weaned rats (PD3-PD10: weaned) and reached the level observed in virgin animals 3 days after removal of pups ( Fig. 2A, upper and lower panels).
Immunostaining with an anti-IRF-1 antibody demonstrated that IRF-1 was expressed predominantly in the epithelial cells of the ducts and lobules of the mammary gland. Expression pattern of the IRF-1 protein coincided with that of the IRF-1 mRNA; expression was detectable in the mammary glands of virgin and early pregnant (G5) animals but not during late pregnancy (G21) or in the lactating glands (Fig. 2B, upper and lower panels). No signal was detected when sections were stained with either affinity-purified rabbit immunoglobulins or with anti-IRF-1 antibody preincubated with the cognate peptide (data not shown).
Since IRF-1 expression in the mammary epithelial cells decreased during pregnancy, a time at which the cells undergo massive proliferation, we wished to determine whether expression of the IRF-1 protein would be lower in breast tumor tissue compared with matched normal glands. We immunostained 23 tumors of various histologic grades with an anti-IRF-1 antibody. Expression of IRF-1 was absent in three tumors but present in the adjacent normal uninvolved ducts and lobules (Fig. 2C, upper panel). Two of these were of poorly differentiated histologic grade, and one tumor was moderately differentiated. In seven patients, staining for IRF-1 was patchy and limited to 20 -80% of the tumor tissue (data not shown). In the other 13 patients expression of IRF-1 was detectable in both the tumor and the surrounding normal tissue (Fig. 2C, lower panel).
Since these results suggest a growth suppressive effect for IRF-1 in breast cancer, we analyzed whether IRF-1 could induce apoptosis of breast cancer cells. MDA-MB-468 cells in logarithmic growth phase were transiently transfected with a construct, which encodes for the IRF-1 protein, and a plasmid encoding the cell surface marker CD20 as described by Ref. 28. Vector-transfected cells were used as controls. Cells were fixed 72 h after transfection and stained for CD20 expression. An- nexin V staining of CD20-positive cells demonstrated that expression of IRF-1 in breast cancer cells led to a ϳ3.5-fold and 5-fold increase in cells in early and late stages of apoptosis, respectively, compared with vector-transfected cells (Fig. 2D).
MIS and Interferon Co-stimulate IRF-1-Since IRF-1 is strongly induced by interferons (15) and MIS, we tested the effect of IFN-␥ on MIS-mediated induction of IRF-1 expression. IFN-␥ induced IRF-1 expression in T47D cells and simultaneous addition of MIS and IFN-␥ further increased the induction of the IRF-1 mRNA in both T47D and MDA-MB-468 cells (Fig.  3A). We had previously demonstrated that MIS induces the DNA binding activity of NFB protein complexes in human mammary epithelial cells, breast cancer cells and in the normal breast in vivo (3,4). In order to determine the molecular mechanism by which MIS induces IRF-1 expression in breast cancer cells, gel shift assays were performed using NFB or STATinducing element (SIE) oligonucleotides containing the relevant DNA binding consensus sequences (Fig. 3B). As reported previously (3,4), MIS induced NFB DNA binding activity consisting of p50 and p65 NFB subunits in T47D cells. Binding to the SIE DNA sequence was not observed suggesting that MIS does not evoke STAT DNA binding in these cells. IFN-␥ however induced SIE DNA binding activity but did not activate the DNA binding activity of NFB. Antibody supershift experiments demonstrated that the STAT-DNA protein complex induced by IFN-␥ contained the STAT-1 protein but not STAT-3 or STAT-5␣. In agreement with these results, Western blot analysis of nuclear and cytoplasmic extracts of MIS and IFN-␥-treated T47D cells demonstrated that unlike IFN-␥, MIS did not induce phosphorylation of the STAT1 protein (Fig. 3C). In order to determine whether activation of the NFB signaling cascade by MIS was responsible for the induction of IRF-1 mRNA, we generated T47D cell clones which express the dominant negative inhibitor of IB (IB␣-DN). In the rat IB␣-DN transgene used in these experiments, two serine residues at positions 32 and 36 are replaced by alanines. Hence the resulting IB␣-DN protein cannot be phosphorylated in response to activation signals. Thus it functions as a super repressor of NFB activation (31). Two T47D cell clones expressing the IB␣-DN transgene were identified by the lack of NFB activation following MIS treatment (Fig. 3D, upper  panel). Induction of IRF-1 by MIS was greatly reduced in the two clones harboring IB␣-DN compared with cells transfected with the empty vector (Fig. 3D, lower panel). Thus MIS-induced IRF-1 requires activation of NFB DNA binding activity. We next tested whether co-stimulation of IRF-1 by MIS and IFN-␥ would be impaired in cells expressing IB␣-DN. Northern blot analysis of IB␣-DN-expressing cells treated with MIS, IFN-␥ or both demonstrated that expression of IB␣-DN did not interfere with IFN-␥-induced IRF-1 expression. The stimulation of IRF-1 mRNA by a combination of MIS and IFN-␥ was equivalent to that induced by IFN-␥ alone since IRF-1 induction by MIS was impaired in these cells (Fig. 3E). Thus it is likely that co-stimulation of IRF-1 by MIS and IFN-␥ in breast cancer cells is mediated through activation of NFB and STAT pathways, respectively. Co-stimulation of IRF-1 was also observed when breast cancer cells were treated with a combination of MIS and IFN-␤, a class I interferon (Fig. 3F).
Synergistic Induction of CEACAM1 by MIS and IFN-␥ Is Mediated by IRF-1-CEACAM1 also known as biliary glycoprotein (BGP) is a Ca 2ϩ -dependent cellular adhesion molecule that is expressed in epithelial cells (32,33). An interferon-sensitive response element (ISRE) in the CEACAM1 promoter is specifically protected by IRF-1 in DNA footprints and is required for induction of a CEACAM1 promoter-driven reporter construct by IRF-1 (34). Both MIS and IFN-␥ induced CEACAM1 expression in T47D cells (Fig. 4A, left panel). Induction of CEACAM1 by these two ligands was also observed in MDA-MD-468 cells (data not shown). Interestingly, simultaneous addition of MIS and IFN-␥ resulted in synergistic induction of CEACAM1 expression (Fig. 4A, right panel).
In order to determine whether induction of CEACAM1 by MIS and IFN-␥ was mediated through IRF-1, we generated T47D cells that stably express the antisense IRF-1 transcript. The ability of antisense IRF-1 to block the translation of IRF-1 protein was demonstrated by transient transfection of sense and antisense IRF-1 constructs into COS cells (Fig. 4B). Antisense IRF-1 expression inhibited the translation of IRF-1 protein derived from an IRF-1 expression construct. Phosphorimager analysis of CEACAM1 mRNA induction in breast cancer cells expressing antisense IRF-1 (Fig. 4C, upper panel) demonstrated that MIS-and IFN-␥-induced CEACAM1 expression by 3-and 8-fold, respectively, in control cells and by 2-and 5-fold, respectively, in two clones expressing the antisense IRF-1 transcript suggesting that antisense IRF-1 slightly lowered CEACAM1 induction by MIS or IFN-␥ (Fig. 4C, lower panels). However, the synergistic up-regulation of CECAM1 mRNA by combined treatment with MIS and IFN-␥ was greatly impaired by the expression of antisense IRF-1; MIS and IFN-␥ together induced CEACAM1 expression by 30-fold in control cells while its induction was additive (6 -7-fold) in both clones expressing antisense-IRF-1 RNA (Fig. 4C, lower panel).
Effect of MIS and IFN-␥ on Breast Cancer Cell Growth-Since the signaling events initiated by MIS and IFN-␥ converge to increase the magnitude of gene expression, we next tested their effect on the growth of breast cancer cells. Treatment of MDA-MB-468 cells with either MIS or IFN-␥ inhibited growth and the presence of both inhibited growth better ( Fig. 5A;  n ϭ 8).
In order to identify the mechanism by which MIS and IFN-␥ inhibit growth, assays to estimate cell cycle progression and apoptosis were performed. MDA-MB-468 cells were treated with MIS, IFN-␥, or MISϩIFN-␥ for 72 h, and the fraction of cells in each phase of the cell cycle was estimated by fluorescence-activated cell sorting (Fig. 5B). Compared with untreated cells, MIS or IFN-␥ treatment consistently led to a statistically significant decrease in the number of cells in the S-phase of the cell cycle (p Ͻ 0.001 by Student's t test). Interestingly, in cultures treated with a combination of MIS and IFN-␥, the percentage of cells in the S-phase did not demonstrate a greater decrease compared with that seen with either agent alone and these cultures did not exhibit any other extensive alteration in cell cycle distribution compared with cells treated with either agent alone. Thus the enhanced inhibition of breast cancer cell growth by MIS and IFN-␥ could not be explained by combined changes in cell cycle progression compared with treatment with either agent alone.
Translocation of annexin V from the inner surface of the plasma membrane to the outside occurs after initiation of apoptosis and thus serves as a marker of apoptosis.   (Fig. 5C). MIS consistently increased apoptosis in several experiments but its effect was much less potent than that of IFN-␥ at the concentration tested. However, treatment of cells with a combination of MISϩIFN-␥ together resulted in a synergistic increase in the fraction of cells in early and late stages of apoptosis. Thus growth inhibition of MDA-MB-468 cells following co-treatment with MIS and IFN-␥ results from enhanced of apoptosis rather than a combinatorial effect on the cell cycle. DISCUSSION MIS is a sexually dimorphic hormone that plays an important role in proper sexual development in male embryos (1). Interferons are antiviral and immunoregulatory proteins, which can negatively regulate growth in various cell types (35). IRF-1 mediates many IFN-␥-induced responses within cells by enhancing gene expression (14,15) and its expression is also modulated by the cytokines TNF-␣, IL-1, IL-6, and prolactin (15). TGF␤ can either up-or down-regulate the expression of IRF-1 depending on its growth regulatory role in a particular cell type. In human embryonic lung fibroblasts, TGF␤-stimulated DNA synthesis was associated with suppression of IRF-1 expression whereas in human cholangiocarcinoma cells, TGF␤ suppressed DNA synthesis through up-regulation of IRF-1 (36). Our results demonstrate that in addition to TGF␤, MIS and activin A also induce IRF-1 suggesting that members of the TGF␤ superfamily may represent another class of molecules that can regulate IRF-1 expression.
Analysis of IRF-1 expression in the rat mammary gland demonstrated a gradual decline in mRNA that begins at the early stages of pregnancy suggesting that it may be a negative regulator of growth and/or differentiation in mammary epithelial cells. The RNA and protein were almost undetectable during late stages of pregnancy and lactation but recovered to levels seen in virgin animals nearly 3 days after removal of pups. However, Chapman et al. (37) analyzing total protein isolated from mammary glands of lactating and weaned mice by Western blot demonstrated that IRF-1 protein was expressed in the lactating mammary glands of mice and that levels did not change significantly during 24, 48, 72, and 96 h of involution. Western blot analysis is a more sensitive analytical tool than immunostaining to detect low levels protein expression. Thus it is possible that the discrepancy between these two observations results from the difference in sensitivity between the two techniques. Alternatively, the difference could also be attributed to the samples analyzed; total IRF-1 expression in the mammary gland (37) versus expression in the epithelial compartment in which IRF-1 protein expression is maintained at very low levels during late pregnancy, lactation, and early stages of involution but is up-regulated at weaning.
Many lines of evidence demonstrate that IRF1 plays a key role in growth control (25). The IRF-1 gene maps to the chromosomal region 5q31.1 that is frequently deleted in human leukemia (38). The tumor suppressor activity of IRF-1 is also suggested by loss of an IRF-1 allele in esophageal and gastric cancer (39 -41). IRF-1 immunostaining of breast cancer specimens demonstrated that the protein was not detectable in 14% of invasive tumors. Expression did not correlate with estrogen receptor status or Page grade but correlated with nuclear grade; it was undetectable in 41% of breast tumors of high nuclear grade (42). Consistent with the results reported by Doherty et al. (42), our results demonstrated loss of IRF-1 expression in 13% (3/23) of breast tumors compared with matched normal control tissue. Furthermore, 7 tumors demonstrated patchy staining in 20 -80% of the tumor tissue. Thus some breast cancers may by-pass the growth-inhibitory effect exerted by IRF-1 by down-regulating its expression. In agreement with this concept, expression of IRF-1 in breast cancer cells results in the robust induction of apoptosis.
Paradoxically, examination of involuting mammary glands of IRF-1-null mice demonstrated accelerated apoptosis compared with wild-type mice at 48 h of involution. However, no difference in morphology was evident in the mammary glands isolated from control and IRF-1-null mice at 72 h of involution (37). These results suggest that IRF-1 may be a suppressor of premature epithelial apoptosis in the mammary gland. Thus it is possible that IRF-1 serves different functions during various stages of postnatal mammary gland development, neoplastic transformation, and tumorigenic process of the breast.
Induction of IRF-1 by IFN-␥ occurs through phosphorylation of the latent transcription factor STAT1, homodimers of which bind to the IRF-1 promoter (15). However, the presence of a putative NFB site within the IRF-1 promoter (43,44) renders it responsive to extracellular signals that activate the NFB pathway. Induction of IRF-1 by MIS in breast cancer cells was mediated by activation of NFB and addition of methylthioadenosine to inhibit STAT1 methylation (45) lowered IFN-␥-induced IRF-1 expression (data not shown). Thus IRF-1 co-stimulation by MIS and IFN-␥ in breast cancer cells may occur through activation of these two pathways.
Several growth regulatory genes including those with antiproliferative activity such as IFN␣/␤, p21, and CEACAM1 have IRF-1 DNA recognition sites in their promoters (25,34). In HeLa and HT-29 cells, IFN-␥ up-regulated a CEACAM1 promoter-driven-luciferase construct by 2-and 2.5-fold, respectively, an effect that was abrogated upon mutating the interferon response element that binds IRF-1 (34). However, this report did not evaluate the effect of IFN-␥ and IFN-␥-induced IRF-1 on the induction of endogenous CEACAM1 mRNA. In breast cancer cells expression of antisense IRF-1, which ablates translation of the IRF-1 protein decreased CEACAM1 induction slightly when MIS and IFN-␥ were used alone suggesting that IRF-1 may be partially responsible for this inductive process. Quantification of band intensities demonstrated that induction of CEACAM1 by a combination of MIS and IFN-␥ was strictly additive in T47D cells expressing antisense IRF-1. However, the synergistic up-regulation of CEACAM1 by MIS and IFN-␥ was completely abrogated in both T47D cell clones stably expressing the antisense IRF-1 transcript suggesting that IRF-1 may be involved in the interaction between MIS and IFN-␥ leading to the synergistic induction of CEACAM1.
IRF-1 has been implicated in mediating the IFN-␥ contribution to synergistic enhancement of transcription in other experimental systems. Enhancer elements that bind IFN-␥-responsive transcription factors including an IRF-1 binding site have been shown to be involved in the synergistic induction of the iNOS promoter-driven luciferase construct by IFN-␥ (46). In IRF-1-null macrophages, the ability of IFN-␥ to up-regulate as well as synergistically induce Cox-2 mRNA expression was abrogated (47). The synergistic induction of transcription by IRF-1 has been shown to depend on protein-protein interaction (48 -50). Further analysis of the CEACAM1 promoter-driven reporter construct may be required to delineate the process by which IRF-1 mediates the synergistic interaction between MIS and IFN-␥ in the induction of the CEACAM1 gene in breast cancer cells.
CEACAM1, located on chromosome 19 (51), is down-regulated in human colon and prostate cancers (52,53) and in about 30% of breast carcinomas (54,55). Consistent with its tumor suppressor function, introduction of CEACAM1 into MDA-MB-468 cells suppressed tumorigenicity in nude mice (56). In normal mammary epithelial cells, CEACAM1 staining is confined to the luminal surface and its localized expression appears to be important in lumen formation (55,57) suggesting that CEACAM1 expression may be important in differentiation of mammary epithelial cells. Furthermore, expression of CEACAM1 in the BGP-negative MCF7 cells induces cell death with occasional formation of acini when grown in extracellular matrix (57). The synergistic up-regulation of CEACAM1 by MIS and IFN-␥ suggests that the level of CEACAM1 expression in the mammary epithelial cells may depend on the integrated response to various extracellular signals received by the cell. Whether the synergistic induction of CEACAM1 by MIS and IFN-␥ can reinitiate the differentiation program in breast cancer cells remains to be determined.
IFN-␥ in combination with IFN-␤ has been shown to induce the regression of human breast cancer cell lines MCF7 and BT20 grown as xenografts in nude mice (58). Although the antitumor effect of IFN-␥ in vivo has been well documented, toxicity associated with exposure to IFN-␥ has diminished its utility in treatment (59). The ability of MIS to augment IFN-␥-induced growth inhibitory/differentiation signals such as CEACAM1 and apoptosis of breast cancer cell growth, suggests that MIS may prove to be beneficial in harnessing the antitumor effects of this cytokine, especially since high levels of MIS have not shown any harmful effects in humans (60).