Involvement of IFN Regulatory Factor (IRF)-1 and IRF-2 in the Formation and Progression of Human Esophageal Cancers

Introduction

Esophageal cancer is an extremely aggressive tumor ranking the eighth most common malignancy and the sixth most frequent cause of cancer death worldwide ( 1). Two main pathologic subtypes are esophageal squamous cell carcinomas (ESCC) and esophageal adenocarcinomas. China reportedly has the highest morbidity and mortality of ESCCs worldwide with a remarkable geographic distribution ( 2). Although the pathologic progression of ESCCs has been well described ( 3), the molecular mechanisms underlying tumorigenesis of ESCCs remain largely unknown. Loss of heterozygosity (LOH) at either single or multiple loci on 5q is frequent, suggesting these cancers harbor altered tumor-suppressor genes in this region. The smallest commonly deleted region is at 5q31.1, the location of the IFN regulatory factor (IRF)-1 gene locus ( 4).

The IRF family is a group of transcription factors, and nine IRF members (IRF-1 to IRF-9) are present in man ( 5). These IRF molecules play pivotal roles in antiviral defense, immune response/modulation, and cell growth regulation by stimulating expression of IFN-α/IFN-β, IFN-stimulated genes, as well as induction of expression of cytokines and chemokines ( 6, 7). IRF-1 is the first identified member of the IRF family; it can be effectively induced in most cell types after exposure to IFN-γ ( 6, 7). When translocated into the nucleus, IRF-1 binds to promoter regions of many IFN-γ–inducible genes and activates their transcription leading to inhibition of cell proliferation and stimulation of cell apoptosis ( 8, 9). IRF-2 is also induced by IFN-γ and acts as an antagonist to IRF-1. IRF-2 can bind to the same DNA sequences as IRF-1 but down-regulates or blocks the transcription of IRF-1 target genes ( 10). IRF-2 is generally synthesized after IRF-1 upon cellular exposure to IFN-γ. Also, IRF-2 has a greater protein stability (8-h half-life) compared with IRF-1 (30-min half-life; refs. 10, 11). IRF-2, therefore, plays a role in the feedback inhibition to the effects of IFN-γ mediated by IRF-1 ( 8, 9). Furthermore, evidence is accumulating to suggest that IRF-1 and IRF-2 have antioncogenic and oncogenic potentials, respectively ( 8– 10). Overexpression of IRF-2 in NIH3T3 cells promotes their transformation and tumorigenicity in nude mice ( 10). In contrast, forced expression of IRF-1 can reverse the IRF-2–mediated transformed phenotype ( 10). IRF-1 overexpression in fibrosarcoma cells (MCA101) can also suppress their malignant phenotype in syngeneic mice ( 12). Similarly, embryonic fibroblasts from IRF-1 knockout mice, but not wild-type mice, were transformed by forced expression of mutant c-Ha-ras oncogene ( 13). The loss of IRF-1 expression and gain of IRF-2 expression in human melanoma and breast cancer cells have been associated with their more malignant phenotype ( 14, 15). In addition, incremental loss of IRF-1 expression paralleled the advancement of hepatic and endometrial cancers ( 16, 17). Accordingly, the expression patterns of IRF-1 and IRF-2 in clinical cancer specimens support IRF-1 as a tumor suppressor and IRF-2 as an oncoprotein.

It has been reported that transfection of IRF-1 in esophageal adenocarcinoma cell lines can induce their apoptosis ( 18), but the expression pattern and function of IRF-1 and IRF-2 in ESCCs are poorly understood. In addition, the significance of the two molecules in clinical diagnosis and prognosis for ESCCs has not been investigated thus far. In the present experiments, therefore, we determined the expression pattern of IRF-1 and IRF-2 in ESCC samples as well as matched normal tissues and correlated these finding with clinical features of ESCCs. Moreover, we explored the function of IRF-1 and IRF-2 in ESCC cell lines, confirming, for the first time, that the ratio of expression of these two proteins is important for development and progression of ESCCs.

Materials and Methods

ESCC tissue samples and cell line. Fifty pairs of primary ESCCs and their corresponding adjacent normal tissues, which were at least 3 to 4 cm away from the cancer, were obtained from ESCC patients treated at First Affiliated Hospital of Zhengzhou University (Henan, China) from 2002 to 2005, after their written informed consent. The clinical information of the patients is summarized in Supplementary Table S1, and none of the patients received any neoadjuvant therapy. Each specimen was divided into two parts: (a) one part was sectioned and examined histologically by traditional H&E staining for the presence of >80% tumor cells (cancer sample) or only normal cells without any inflammation or tumor invasion (matched normal sample) and (b) the other part was frozen in liquid nitrogen and stored at −80°C until analysis. Our work was approved by the Institutional Review Board of the Institute for Nutritional Sciences, Chinese Academy of Sciences. ESCC cell line EC109 (Cell Bank of Type Culture Collection of Chinese Academy of Sciences, Shanghai Institute of Cell Biology, Chinese Academy of Sciences; refs. 19, 20) was cultured in RPMI 1640, supplemented with 10% fetal bovine serum, 10 units/mL penicillin, and 10 units/mL streptomycin, at 37°C in a humidified atmosphere containing 5% CO₂.

Real-time PCR analysis. Total RNA was isolated from patient tissues and ESCC cell line using TRIzol reagent (Invitrogen, Carlsbad, CA) according to the standard protocol. Two micrograms of high-quality RNAs were processed directly to cDNA by reverse transcription with Superscript II (Invitrogen) following the manufacturer's instruction in a total volume of 50 μL. Primers for the genes tested in the present experiments were designed using software PRIMER3 ⁵ (Supplementary Table S2). Amplification reactions were done in 20 μL of the LightCycler-DNA Master SYBR Green I mix (Roche Applied Science, Penzberg, Germany) with 10 pmol primer, 2 mmol/L MgCl₂, 200 μmol/L deoxynucleotide triphosphate mixture, 0.5 units Taq DNA polymerase, and universal buffer. All of the reactions were done in triplicate in an iCycler iQ system (Bio-Rad, Hercules, CA), and the thermal cycling conditions were as follows: 95°C for 3 min; 40 cycles of 95°C for 30 s, 58°C for 20 s, and 72°C for 30 s; 72°C for 10 min.

The relative mRNA levels of target genes to that of β-actin in either clinical samples or cultured cells were calculated according to the methods described by Xie et al. ( 21) with some modifications. In brief, reactions were characterized at the point during cycling when amplification of the PCR product was first detected after a fixed number of cycles, which is defined as Ct. The target message in unknown samples was quantified by measuring the Ct value, and the Ct value of β-actin was also measured as the endogenous RNA control. The levels of target genes in each sample were normalized on the basis of its β-actin content through the formula: Level_target/Level_β-actin = 2^{(Ct_β-actin − Ct_target)}. Spearman's rank correlation coefficient was used to evaluate the relationships between the mRNA expression of either IRF-1 or IRF-2 with clinicopathologic features of ESCC samples. Paired t test was adopted to study the expressions of target genes in ESCC cell line. The statistical results were considered significant at P < 0.05 and highly significant at P < 0.01. All data analyses were done using the program SPSS for Windows.

Immunohistochemistry and scoring system. For immunohistochemistry, ESCCs and matched normal esophageal tissues were frozen in a cryostat chamber and 10 μm sections were collected on glass slides. The sections were fixed in ice-cold acetone for 30 min, washed in 0.01 mol/L PBS (8 mmol/L Na₂HPO₄, 2 mmol/L NaH₂PO₄, and 150 mmol/L NaCl) for 3 × 5 min, blocked for 1 h in 0.01 mol/L PBS supplemented with 0.3% Triton X-100 and 5% normal goat serum, and then incubated with either IRF-1 (1:500) or IRF-2 (1:500) at 4°C overnight. After brief washes in 0.01mol/L PBS, sections were incubated for 2 h in 0.01 mol/L PBS with horseradish peroxidase–conjugated goat anti-rabbit IgG (1:1,000), followed by development with 0.003% H₂O₂ and 0.03% 3,3′-diaminobenzidine in 0.05 mol/L Tris-HCl (pH 7.6). Immunohistochemistry for each sample was done at least thrice, and all sections were counterstained with hematoxylin. Negative controls consisted of substitution of the primary antibody with normal rabbit serum at the same dilution (Supplementary Fig. S1).

The immunohistochemical evaluation for intratumor expression of IRF-1 was carried out independently by three pathologists blinded to the patient's clinical information, and a scoring system was developed as follows. The cells bearing obvious brown signals of either IRF-1 or IRF-2 compared with negative controls were considered positive. The intensity of staining was rated as either 0 (no signal as the negative controls), 1 (weak), or 2 (strong); the percentage of positive tumor cells was graded as 0 (no cells), 1 (1–25% of total tumor cells), 2 (26–50%), 3 (51–75%), and 4 (75–100%). The immunoreactive score for whole slides was calculated by multiplying the score of percentage positive cells and the score of staining intensity. As a result, seven grades were scored as 0, 1, 2, 3, 4, 6, and 8. The tumors with scores ≥2 were referred as positive, and those with scores <2 were negative. Relationships between intratumor expression of either IRF-1 or IRF-2 protein with clinicopathologic features were explored by Spearman's rank correlation coefficient and done using the program SPSS for Windows.

Construction and transfection of plasmids. Full length cDNA of either IRF-1 or IRF-2 was subcloned from vectors of pCMV–IRF-1 and pCMV–IRF-2, respectively, which were kindly provided by Dr. W.A. Chow ( 22), and the products of IRF-1 or IRF-2 were inserted into pcDNA3.1 vectors at appropriate sites (Invitrogen). The IRF-1–pcDNA3.1, IRF-2–pcDNA3.1, and empty pcDNA3.1 plasmids were then transfected into EC109 using LipofectAMINE 2000 reagent (Invitrogen). The transfected cells were selected in the presence of G418 (Invitrogen) (600 μg/mL), and resistant clones were further confirmed by Western blotting (see supplemental Materials and Methods).

RNA interference of IRF-2. Target sequence for IRF-2–small interfering RNA was GGACCAACAAGGGCAGTGG, and the disturbed nucleotide sequence of IRF-2–small interfering RNA was GAATGCGAGCGAGCGAGCA as a negative control. FG12 lentiviral vector was used to produce small double-stranded RNA (small interfering RNA) to inhibit target gene expression in ESCC cells. And the information and usage on this vector system has been previously described in detail by Qin et al. ( 23).

Tumorigenicity assay in vivo. Female nude mice were housed under standard conditions. The animal protocols were done in agreement with SIBS Guide for the Care and Use of Laboratory Animals and approved by Animal Care and Use Committee, Shanghai Institutes for Biological Sciences. Six-week-old female nude mice were s.c. injected at two sites in the flanks with 1 × 105 ESCC cells per flank. The resulting tumors were measured with calipers every 4 days, and tumor volume (cm3) was calculated using the standard formula: length × width × height × 0.5236 ( 24). Four weeks after injection, tumors were harvested from ether-anesthetized mice. Data were presented as tumor volume (mean ± SD). Statistical analysis was done using the Student's t test by the program SPSS for Windows.

The information about antibodies/reagents and the methods of some other experiments were compiled in the supplemental “Materials and Methods”.

Results

Expressions of IRF-1 and IRF-2 mRNA in ESCCs and matched normal esophageal tissues. Real-time PCR was initially done to evaluate the levels of IRF-1 and IRF-2 mRNA in 50 pairs of ESCCs and matched normal esophageal tissues. The mRNA level of β-actin in each sample was also quantified as an internal standard and used to normalize the levels of IRF-1 and IRF-2 from the same sample. The expression profiles of IRF-1 and IRF-2 mRNA were determined in panels of 50 tumor/normal pairs (Supplementary Fig. S2). For IRF-1, 12 pairs (24%) showed <2-fold change in IRF-1 expression, 11 pairs (22%) had >2-fold increase of IRF-1 expression in the tumors, and 27 pairs (54%) had > 2-fold decrease of IRF-1 expression in the tumors. For IRF-2, 10 pairs (20%) displayed <2-fold change in IRF-2 expression, 35 pairs (70%) had >2-fold increase of IRF-2 expression in the tumors, and 5 pairs (10%) had >2-fold decrease of IRF-2 expression in the tumors. In ESCCs, therefore, the data showed that IRF-1 was often decreased and IRF-2 was frequently increased compared with matched normal tissues.

Relationships between mRNA expression of either IRF-1 or IRF-2 and ESCC clinicopathologic features. Because our results showed that the down-regulation of IRF-1 mRNA and the up-regulation of IRF-2 mRNA were frequently detected in the ESCC samples, the correlations between either IRF-1 mRNA down-regulation or IRF-2 mRNA up-regulation and ESCC clinicopathologic features were estimated using Spearman's rank correlation coefficient. Statistical analysis revealed that the up-regulation of IRF-2 mRNA in ESCCs compared with matched normal esophageal tissues was positively correlated with tumor stage (P = 0.001), depth of tumor infiltration (P = 0.006), lymph node metastasis (P = 0.015), and tumor size (P = 0.011; Table 1 ), disclosing a close correlation between IRF-2 mRNA up-regulation and ESCC progression. However, no correlations existed between IRF-2 mRNA up-regulation and tumor differentiation, lymph node metastasis number, patient age, or gender. Although real-time PCR results showed that IRF-1 mRNA down-regulation was frequently observed in ESCCs compared with matched normal esophageal tissues, no statistical relationships were noted between IRF-1 mRNA down-regulation with the clinicopathologic features except lymph node metastasis number (P = 0.017; Table 1), which implied that low level of IRF-1 might favor metastasis of ESCC to lymph nodes.

Expressions of IRF-1 and IRF-2 protein in ESCCs and matched normal esophageal tissues. The in vivo distribution of IRF-1 and IRF-2 proteins was also investigated by immunohistochemistry. Representative figures showed that IRF-1 was expressed in most of the normal esophageal epithelial cells with the distribution in all layers of the esophageal mucosa (i.e., superficial layer, epibasal layer, and basal layer; Fig. 1A ). The local higher magnification further showed that the staining of IRF-1 was mainly concentrated in the nucleus ( Fig. 1A'). In the cancerous counterpart, however, the intensity of IRF-1 was attenuated ( Fig. 1B) and the distribution of IRF-1 was mainly in the cytoplasm, not in the nucleus ( Fig. 1B). In normal esophageal tissues, IRF-2 was only seen in a few cells ( Fig. 1C) with the positive immunoreaction in the nucleus ( Fig. 1C), and the sparse IRF-2–positive cells seemed to appear mainly near the basal layer of the esophageal mucosa, whereas the expression of IRF-2 was abundant in the corresponding ESCC tissues ( Fig. 1D) and observed nearly in all tumor cells with intense staining in both the cytoplasm and nucleus ( Fig. 1D).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Immunohistochemical analyses of IRF-1 and IRF-2 expression in ESCCs and matched normal esophageal tissues. In noncancerous tissues (A), intense IRF-1 immunoreactivity (dark gray) is observed, and the IRF-1–positive cells are distributed in all layers of the esophageal mucosa: superficial layer (SL), epibasal layer (EBL), and basal layer (BL). The greater magnification in the smaller frame in (A) further shows that the signal of IRF-1 is strong in the nucleus and also visible in the cytoplasm (A′). In corresponding ESCC tissues, however, IRF-1 expression is predominantly weaker (B), and the protein is only found in the cytoplasm (B′). Unlike IRF-1, IRF-2 is only seen in a few cells (C) in normal esophageal tissues with the positive immunoreaction in the nucleus (C′), and the sparse IRF-2–positive cells seemed to appear mainly near the basal layer of the esophageal mucosa. In the corresponding ESCC tissues, however, IRF-2 is prominently expressed in nearly all tumor cells (D) with intense staining in both the cytoplasm and nucleus (D′). All sections were counterstained with hematoxylin. T, tumor mass. Bar, 50 μm for (A–D); 20 μm (A′–D′).

Relationships between intratumor expression of either IRF-1 or IRF-2 and ESCC clinicopathologic features. All immunohistochemical slides were scored according to the description in Materials and Methods, and classified as “positive” or “negative.” Relationship between the intratumor expression of either IRF-1 or IRF-2 and clinicopathologic features was explored using Spearman's rank correlation coefficient. Similar to the statistical analysis of mRNA, IRF-2 protein was also positively correlated with tumor stage (P = 0.006), depth of tumor infiltration (P = 0.003), lymph node metastasis (P = 0.042), and tumor size (P = 0.006; Table 2 ). No correlations existed between IRF-2 protein with tumor differentiation, lymph node metastasis number, patient age, or gender. Although real-time PCR results showed that IRF-1 mRNA levels were significantly down-regulated in ESCC compared with matched normal esophageal tissues and the down-regulation of IRF-1 mRNA was correlated with the lymph node metastasis number in ESCCs, no statistical relationships were displayed between intratumor expression of IRF-1 protein and the clinicopathologic features of the ESCC samples.

Effects of either IRF-1 or IRF-2 overexpression on ESCC cells in vitro. Analysis of the clinical samples disclosed the potential involvement of IRF-1 and IRF-2 in the development and progression of ESCCs, but the mechanism underlying these events is poorly understood, motivating us to investigate the functional role of IRF-1 and IRF-2 in ESCC cells. The ESCC cell line EC109 ( 19, 20) was used. Western blotting showed that IRF-2 was constitutively expressed in wild-type EC109 (109/WT) but IRF-1 was not detected ( Fig. 2A ). By stable transfection of 109/WT with either IRF-1 or IRF-2, we established five cell lines having either low expression of IRF-1 (109/1L), high expression of IRF-1 (109/1H), low expression of IRF-2 (109/2L), high expression of IRF-2 (109/2H), or vector only (109/V) as control ( Fig. 2A). Because p21^WAF1/CIP1 and histone H4 are well-characterized target genes for IRF-1 and IRF-2, respectively ( 25, 26), their expressions were tested by either Western blotting or real-time PCR. Levels of histone H4 mRNA were clearly increased in both 109/2L and 109/2H ( Fig. 2A and B), whereas up-regulation of p21^WAF1/CIP1 was only observed in 109/1H although both 109/1H and 109/1L expressed IRF-1 ( Fig. 2A).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Effects of IRF-1 and IRF-2 on EC109 cells. A, Western blotting indicates that IRF-2 is detected in 109/WT but IRF-1 is not. 109/WT is stably transfected with either IRF-1 or IRF-2 expression vector, and five clones are selected for additional experiments: 109/1L, 109/1H, 109/2L, 109/2H, and 109/V as control. As a target gene of IRF-1, p21^WAF1/CIP1 is examined by Western blotting to identify the validity of exogenous IRF-1. Protein expression of cyclin-D1, total Rb, phosphorylated Rb, Bcl-2, and casepase-9 in 109/WT, 109/V, 109/2L, 109/2H, 109/1L, and 109/1H are also tested. Caspase-9 antibody recognizes its precursor and cleaved peptide (p35). B, the expression of histone H4, a target gene of IRF-2, is examined by real-time PCR, validating the efficiency of exogenous IRF-2, and the mRNA expressions of cyclin-D1, Rb, Bcl-2, and caspase-9 are also measured by real-time PCR. **, P < 0.01 versus 109/WT; Student's t test. Bars, SD.

Further studies showed that proliferation rates of 109/2L and 109/2H were faster than that of 109/WT. In contrast, growth of 109/1H was retarded, and no change in cell number was found for 109/1L (Supplementary Fig. S3A). Fluorescence-activated cell sorting analysis based on propidium iodide staining revealed obvious difference in spontaneous apoptosis of these clones. 109/WT had ∼6% spontaneous apoptosis, whereas IRF-2–overexpressed clone had decreased apoptosis (∼3% for 109/2L and 2% for 109/2H) compared with 109/WT. In contrast, the 109/1H clone had increased spontaneous apoptosis of EC109 (∼22%), but 109/1L cells showed a similar rate of apoptosis (∼6%) as 109/WT (Supplementary Fig. S3B). Moreover, 109/2H and 109/2L formed more numerous and bigger colonies than 109/WT in soft agar (Supplementary Figs. S3C and S2D). In comparison, clonogenic capacity of 109/1H cells was nearly lost, but the number and size of 109/1L colonies were similar to that of 109/WT (Supplementary Fig. S3C and D). Taken together, forced expression of IRF-2 effectively enhanced the in vitro growth of ESCC cells, which was consistent with our statistical analysis of the clinical samples showing that IRF-2 was positively correlated with ESCC progression.

Regulation of proliferation- and apoptosis-related molecules by IRF-1 and IRF-2 in ESCC cells. Because overexpression of either IRF-1 or IRF-2 prominently affected the proliferation and spontaneous apoptosis of ESCC cells, several proliferation and apoptosis-related molecules were examined in these cells. Up-regulation of cyclin-D1 mRNA and protein was observed in 109/2L and 109/2H compared with 109/WT, and the levels paralleled those of IRF-2. Also, phosphorylated Rb increased in 109/2L and 109/2H cells compared with 109/WT ( Fig. 2A). In contrast, expression of cyclin-D1 (both mRNA and protein) and phosphorylated Rb decreased compared with those in 109/WT ( Fig. 2A). Notably, mRNA and protein levels of total Rb did not vary among these clones ( Fig. 2B). Overexpression of IRF-2 and IRF-1 also altered the levels of several apoptosis-related molecules. Both Bcl-2 mRNA and protein increased proportional to IRF-2 levels of 109/2L and 109/2H compared with 109/WT ( Fig. 2), and the cleavage of caspase-9 (measured as expression of the cleavage product p35) was attenuated in 109/2L and eliminated in 109/2H ( Fig. 2A). In contradistinction, Bcl-2 mRNA and protein levels decreased and cleaved caspase-9 increased in 109/1H ( Fig. 2). Similar to the findings regarding Rb levels, the expression of caspase-9 mRNA was not altered among all cell clones ( Fig. 2B). Furthermore, no variations of cyclin-D1, Rb, Bcl-2, and caspase-9 were observed in 109/1L cells ( Fig. 2). Several other proliferation- and apoptosis-related proteins (e.g., p53, p27^KIP1, cyclin-A1, and Bcl-XL) were also examined, and their levels did not change with forced expression of either IRF-1 or IRF-2 (data not shown).

Cellular localization and activity of IRF-1 and IRF-2 in ESCC cells in vitro. Immunocytochemistry showed that IRF-2 was widely distributed in the cytoplasm and nucleus of 109/WT, 109/V, 109/2L and 109/2H, although little IRF-1 expression was detected ( Fig. 3A ). Interestingly, although IRF-1 in 109/1L was obviously increased in comparison with 109/WT, it was mainly localized in the cytoplasm and IRF-2 remained in both the nucleus and cytoplasm ( Fig. 3A). In 109/1H, IRF-1 was detected throughout the cells, including the cytoplasm and nucleus, but IRF-2 was mainly found in the cytoplasm ( Fig. 3A). Because translocation into the nucleus is a prerequisite for either IRF-1 or IRF-2 to function ( 8, 9), these results, combined with the growth and Western blot studies, suggested that IRF-2 was functional in all cell clones except for 109/1H, whereas IRF-1 was functional in 109/1H but nonfunctional in 109/1L cells.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Immunofluorescence studies on ESCC cells using IRF-1 and IRF-2 antibodies. A, IRF-1 signal is barely visible in 109/WT, 109/V, 109/2L, and 109/2H cells. In 109/1L, IRF-1 is distributed mainly in the cytoplasm, but it is dispersed throughout the cells in the 109/1H clone. Staining for IRF-2 is intense throughout the entire cell of 109/WT, 109/V, 109/2L, 109/2H, and 109/1L clones, but mainly in the cytoplasm of 109/1H cells. B, in the control 109/1L cells, IRF-1 is mainly expressed in the cytoplasm, and IRF-2 is found in both the cytoplasm and nucleus. However, IRF-2–small interfering RNA (IRF-2–siRNA) induces IRF-1 to translocate into the nucleus, accompanied by translocation of the attenuated IRF-2 signal from the nucleus to the cytoplasm. Control small interfering RNA (Con-siRNA) does not change the distribution or expression level of either IRF-1 or IRF-2. C, Western blotting further shows the effective inhibition of IRF-2 by IRF-2–small interfering RNA, which results in the up-regulation of expression of p21^WAF1/CIP1 and the down-regulation of levels of cyclin-D1 and Bcl-2. Meanwhile, phosphorylated Rb is decreased in spite of the stable level of total Rb, and activated caspase-9 (p35) is increased after IRF-2 silence.

Because IRF-2 may compete with IRF-1 for common DNA binding sequences ( 27), we wondered whether the relative concentrations of IRF-1 and IRF-2 could regulate their cellular localization. To pursue this hypothesis, endogenous IRF-2 was decreased in 109/1L using IRF-2–small interfering RNA followed by immunofluorescence analysis. IRF-1 was found largely translocated into the nucleus when IRF-2 was effectively silenced ( Fig. 3B). Meanwhile, expression of p21^WAF1/CIP1 was up-regulated ( Fig. 3C), and levels of cyclin-D1 and Bcl-2 were down-regulated accompanied with the decrease of phosphorylated Rb and the increase of activated caspase-9 ( Fig. 3C). As a result, cell proliferation slowed (Supplementary Fig. S4A), spontaneous apoptosis increased (Supplementary Fig. S4B), and anchorage-independent growth was inhibited (Supplementary Fig. S4C). Therefore, down-regulation of IRF-2 expression enhanced the activity of IRF-1 and effectively depressed the malignant phenotype of ESCC cells.

Effect of forced expression of either IRF-1 or IRF-2 on tumorigenicity of ESCC cells in nude mice. Our in vitro studies indicated that functional overexpression of either IRF-2 or IRF-1 made ESCC cells phenotypically more or less malignant, respectively. Therefore, we also evaluated the effect of overexpression of either IRF-1 or IRF-2 on tumorigenicity of ESCC cells in nude mice. Each animal was injected at two sites in the flanks with the left for 109/1H or 109/2H and the right for 109/V ( Fig. 4A and B ), and the tumor growth was measured every 4 days. The 109/1H cells formed much smaller tumors compared with 109/V (P = 0.004; Fig. 4A and C), whereas 109/2H cells developed larger tumors than control 109/V cells (P = 0.02; Fig. 4B and C).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

Effects of overexpression of either IRF-1 or IRF-2 on tumorigenicity of ESCC cells in nude mice. Xenografts grown in nude mice for 4 wks after the animals were injected with 109/1H (left) and 109/V (right; A) or 109/2H (left) and 109/V (right; B). C, tumor volumes are measured every 4 d. Points, mean volume; bars, SD. *, P < 0.05; **, P < 0.01 versus 109/V; Student's t test.

Relationship between IRF-1 or IRF-2 overexpression and resistance of the ESCC cells to antitumor cytokines. During esophageal metaplasia, immune lymphocytes are activated and produce many cytokines that have a strong antiproliferation activity, but ESCC cells can develop resistance to these cytokines ( 28, 29). Because IRF-1 and IRF-2 are implicated in cytokine-response pathways ( 6– 9), we exposed 109/WT, 109/V, 109/1L, 109/1H, 109/2L, and 109/2H to IFN-α, IFN-γ, or tumor necrosis factor (TNF)-α and measured their proliferation responses to each cytokine. IFN-α, IFN-γ, and TNF-α markedly inhibited the proliferation of IRF-1–expressing ESCC cells. Both IFN-γ and TNF-α moderately inhibited growth of 109/WT and 109/V, whereas IFN-α modestly inhibited proliferation of these cells. In contrast, 109/2L and 109/2H were resistant to the growth inhibitory activity of IFN-α, IFN-γ, and TNF-α (Supplementary Fig. S5). These results suggested that the initial levels of IRF-1 and IRF-2 were pivotal to determine if they would be resistant against these antitumor cytokines.

Discussion

The present study found a reduction of IRF-1 and an elevation of IRF-2 in ESCC tissues compared with their matched normal esophageal samples. Statistical analysis also revealed a positive relationship between IRF-2 and a variety of important clinicopathologic features. Furthermore, forced expression of IRF-1 or silencing of IRF-2 by small interfering RNA in ESCC cell lines effectively inhibited liquid culture proliferation and anchorage-independent clonogenic growth in soft agar, as well as enhanced spontaneous apoptosis. These observations reflect the ability of a relatively high ratio of IRF-1 compared with IRF-2 to attenuate the tumorigenicity of these cancer cells. In contrast, elevated levels of IRF-2 compared with IRF-1 dramatically enhanced the tumorigenicity of the ESCC cells. Moreover, overexpression of either IRF-1 or IRF-2 inhibited or enhanced, respectively, the tumorigenicity of ESCC cells in nude mice. These results are consistent with IRF-1 behaving as a tumor suppressor and IRF-2 as an oncoprotein in ESCCs.

IRF-2 has been shown to promote cell proliferation; and one of the most important mechanisms mediating its effect is the direct stimulation of transcription of histone H4 ( 9). Histone H4 is a phylogenetically conserved cell cycle regulatory element and increases at the G₁-S phase transition to promote DNA replication ( 30, 31). We found up-regulation of levels of histone H4 mRNA after IRF-2 overexpression in the ESCC cell line, consistent with IRF-2 stimulating proliferation of ESCC cells.

Also in our experiments, IRF-2 stimulated expression of cyclin-D1 and phosphorylated Rb. As a tumor suppressor, unphosphorylated Rb prevents transcription of critical cell proliferation–related genes ( 32, 33). Cyclin-D1 stimulates cell growth by inducing Rb phosphorylation that leads to inactivation of Rb and acceleration of the cell cycle ( 34). IRF-2 was also found in our studies to induce the up-regulation of Bcl-2 and the down-regulation of activated caspase-9. Bcl-2 is an important antiapoptotic protein belonging to the Bcl-2 family of proteins ( 35, 36) and is able to impede activation of caspase-9, one of the initial event for caspase-mediated apoptosis ( 37, 38). Therefore, cyclin-D1 and Bcl-2 might be involved in the progrowth and antiapoptosis effects of IRF-2 on ESCC cells.

Additionally, cyclin-D1 or Bcl-2 have been shown to be highly expressed in esophageal cancers ( 39– 41); and thus, IRF-2 might be an upstream stimulator of these proteins in situ. In contrast, forced expression of IRF-1 led to a decrease of cyclin-D1, Bcl-2, and phosphorylated Rb and to an increase of activated caspase-9, associated with decreased cell proliferation and increased apoptosis. Before our study, investigations had not directly linked the effect of either IRF-1 or IRF-2 on expression of either cyclin-D1 or Bcl-2. Investigations have shown that the levels of cyclin-D1 ( 42, 43) and Bcl-2 ( 44, 45) are regulated by exposure of cells to IFN-γ. Because the IFN-γ–induced pathway is largely mediated by IRF-1 and IRF-2, stimulation of cyclin-D1 and Bcl-2 may reflect control by IRF-1 and IRF-2.

In addition, forced expression of IRF-1 induced the expression of p21^WAF1/CIP1, which has been proved to be an executor for inhibition of proliferation of various cells by IRF-1 ( 9, 46). P21^WAF1/CIP1, a cyclin-dependent kinase inhibitor, interacts with the complexes of cyclin/cyclin-dependent kinase and interrupts their activities, causing cell-cycle arrest in the G₁ and/or G₂ phase ( 47, 48). Absence of p21^WAF1/CIP1 generally leads to enhanced cell growth and is observed in many types of cancers, including ESCCs ( 47, 48). Accordingly, the loss of IRF-1 might be a cause of down-regulation of p21^WAF1/CIP1 in ESCCs.

Immunohistochemistry showed that, in ESCCs, IRF-2 was present in the nucleus and in the cytoplasm, although any detectable IRF-1 was localized mainly in the cytoplasm. Because IRF-1 and IRF-2 are both transcription factors, they must translocate into the nucleus for activity ( 8, 9). Consistently, our in vitro experiments showed that when IRF-1 was mainly in the cytoplasm and IRF-2 was intense in the nucleus, transcription of histone H4 was up-regulated but levels of p21^WAF1/CIP1 were not affected, resulting in the stimulation of cell growth. In contrast, when IRF-1 entered the nucleus, accompanied by exclusion of IRF-2 from the nucleus, p21^WAF1/CIP1 expression was markedly induced and cell growth was inhibited. Hence, nuclear localization of either IRF-1 or IRF-2 was well correlated with their functional activation. In addition, silencing IRF-2 by small interfering RNA resulted in nuclear translocation and activation of IRF-1. Taken together, the ratio of IRF-1 to IRF-2 might be the decisive factor determining which of these proteins enters the nucleus and influences the behavior of ESCC cells. Studies have shown that the ratio of IRF-1/IRF-2 oscillates during the cell cycle, attaining its highest level in the growth-arrested cells and its lowest level after growth stimulation; the fluctuation of the ratio of these two proteins assures precise control of growth of normal cells ( 49). In ESCC cells, however, the high levels of IRF-2 probably disrupted the balance between IRF-1 and IRF-2, leading to the dysregulation of cell growth and enhancing the development and progression of ESCCs.

Although ESCC cells are often immunogenic, they are poorly recognized and eliminated by the immune system. Several mechanisms underlying immune privilege of tumor cells have been presented, one of which is that ESCC cells are resistant to antitumor cytokines secreted by activated lymphocytes ( 28, 29). Our results suggested that overexpression of IRF-1 and IRF-2 respectively enhanced and protected ESCC cells from growth suppression mediated by these cytokines. As a result, up-regulation of IRF-2 might promote the immune privilege of ESCC cells during development and progression of the cancer.

In summary, our results in vitro showed that IRF-2 acted as an important oncoprotein in ESCCs, and its elevated expression was correlated with tumorigenicity of ESCC cells in vivo and in vitro and with clinicopathologic features of ESCC samples. In contrast, forced expression of IRF-1 inhibited growth of ESCC cells, indicating it was a potential tumor suppressor of the esophagus. Previous research has described the LOH of the IRF-1 gene in ESCCs, resulting in loss of expression of IRF-1. Our studies suggested that high expression of IRF-2 could also result in loss of function of IRF-1 in ESCCs. Therefore, loss of expression or loss of function of IRF-1 as a result of IRF-2 elevation might play a pivotal role in the development and progression of ESCCs, and measurement of both proteins might be useful indicators in the clinical diagnosis of ESCCs. Our work also hinted that increasing IRF-1 and/or decreasing IRF-2 levels could be a therapeutic approach for ESCCs. Altogether, our results give insights into the function of IRF-1 and IRF-2 in the development and progression of ESCCs and provide novel targets for therapy of this malignancy.