A Differential Gene Expression Profile Reveals Overexpression of RUNX1/AML1 in Invasive Endometrioid Carcinoma

Endometrial carcinoma is among the most common gynecological malignancies in industrialized countries (1). Two clinicopathological types of endometrial carcinoma have been described: type I, endometrioid carcinoma of the endometrium (EEC) and type II, non-endometrioid carcinoma (NEEC). EEC tumors are low-grade and estrogen-related, and they usually develop in peri- and early postmenopausal women. They coexist with or are preceded by complex and atypical endometrial hyperplasia. They usually have a favorable prognosis when treated with appropriate therapy. In contrast, NEECs are high-grade tumors, unrelated to estrogen exposure, which usually develop in older, postmenopausal women and which are not preceded by complex and atypical endometrial hyperplasia (2). NEECs usually present metastasis at the time of diagnosis and often recur, despite aggressive treatment. Among histologic types, the most common endometrial carcinoma corresponds to endometrioid carcinoma of the uterine corpus (EEC) with an incidence of 80% (3).

Some of the molecular events that occur during the development of endometrial carcinoma have been characterized in part, but its molecular pathogenesis is still poorly understood. Among the few genetic alterations that have been recently identified in EEC (3, 4, 5), we should consider (a) microsatellite instability (MSI), frequently associated with mismatch repair deficiency, mutations, and hypermethylation on the MLH1, MSH2, and MSH6 mismatch repair genes (6, 7, 8); (b) mutations in the tumoral suppressor gene, PTEN, reported to be involved in the regulation of focal adhesion, cellular migration, and tumor cell proliferation (9, 10); (c) mutations in the k-RAS oncogene, involved in cellular proliferation and closely associated with MSI and hypermethylation (11, 12); and (d) mutations in the β-catenin gene, which have been related to tumor progression (13, 14). On the other hand, NEEC tumors often present mutations in the p53 gene (14, 15), as well as HER-2/neu amplification (16), and the loss of heterozygosity, reported for a large number of chromosomes, including 3p, 10q, 17p and 18p (17, 18). Nevertheless, these molecular alterations are not present in most cases.

In view of the above inconclusive findings and the poor correlation between genotype and phenotype in EEC, we attempted, in the present work, to identify new genes, which could trigger cell transformation in EEC. For this purpose, we analyzed the differential gene expression profile between tumoral and nontumoral endometrial specimens by using cDNA array hybridization. Among the 53 genes, for which expression was found altered in EEC, the acute myeloid leukemia proto-oncogene RUNX1/AML1 was the most highly up-regulated. The gene expression levels of RUNX1/AML1 were quantified by real-time quantitative PCR and were characterized by tissue array immunohistochemistry. Our results suggest that RUNX1/AML1 plays a role in the process of myometrial infiltration.

In this study, we included a group of patients who underwent surgery for endometrial carcinoma at the Departments of Gynecological Oncology at both the Hospital Vall d’Hebron and the Hospital del Mar in Barcelona, Spain (Tables 1 and 2). Patients ranged from 50 to 80 years of age. None of the patients had received radiation and/or hormonal therapy or chemotherapy treatment before surgery. The protocol was previously approved by the Institutional Review Boards, and informed consent was obtained from all of the patients involved in the study. Each tissue sample was stored at −80°C and was paraffin-embedded until analysis.

We collected nonaffected endometrial epithelium (atrophic endometrium) from the same patients. We also collected endometrial curettage samples from healthy premenopausal women (normal proliferative endometrium). Briefly, healthy endometrial curettages of late proliferative endometrium were cultured in DMEM-10% fetal bovine serum (Life Technologies, Inc., Paisley, United Kingdom) at 37°C in an atmosphere of 5% CO2. Primary cultures allowed epithelial cells to proliferate and reduce the stromal fraction to the minimum. Cell cultures were then tested by fluorescent immunocytochemistry assays for CK19 epithelial antigen (Abcam, Cambridge, United Kingdom), at 1:100 dilutions, as a control for epithelial cells.

Total RNA was extracted with the RNeasy mini kit (Qiagen, Hilden, Germany), following the instructions provided by the manufacturer. For the cDNA array hybridization, mRNA from the total RNA was purified with the Oligotex Kit (Qiagen). The RNA Nano Lab Chip kit (Agilent, Palo Alto, CA) was used to determine the purity of the isolated total RNA. To determine the integrity of the mRNA, Northern blot analysis was performed, according to standard procedures (19), with a 4-kb KIAA0675 probe.

RNAs for hybridization were prepared under the following conditions: (a) atrophic endometrium: a pool of 16 mRNA samples from atrophic endometrium, taken from postmenopausal women and collected at the time of surgery, was generated to obtain gene expression levels in an atrophic context; (b) proliferative endometrium: a normal endometrial pool was generated from four primary cultures taken from proliferate, healthy endometrium, to assess expression levels in a proliferative context; and (c) four dilutions of mRNA, obtained from EECs, as previously described (see details of samples in Tables 1 and 2) were selected.

The 10 × 21 cm nylon membrane cDNA array chip LifeGrid (Incyte Genomics, Inc., Palo Alto, CA) contains 8,345 clones that have been printed in duplicate to assess reproducibility. Messenger RNA (0.5 μg) from the sample collection and 0.5 μg of mRNA from the Universal Human Reference RNA (LifeGrid 1.0.Kit, Incyte Genomics, Inc.), used as a hybridization control, were used to amplify double-strand cDNA. Synthesis was accomplished by oligo(dT) primer, with M-MLVRT retrotranscriptase (200 units/mL; Incyte Genomics, Inc.) at 42°C for 2 hours. The cDNA from endometrial samples were radioactively labeled, including P33 dCTP (2,000–4,000 Ci/mmol) in the reaction mix (New England Nuclear, Boston, MA). Two microliters of each sample were analyzed in a liquid scintillation counter 1410 (Wallac, Turku, Finland) to assess the radioactive incorporation within the probe (up to 25%). Probes were then denatured for 3 minutes at 95°C. Hybridization was performed in 4× SSC, 1× bovine serum albumin, 2 μg/mL DNAs, and 0.1% SDS at 42°C for 16 hours. Slides were washed twice for 30 minutes at 68°C with 2× SSC, 1% SDS and 2× SSC, 0.6% SDS (Sigma, Steinheim, Germany) and were dried and then scanned in a PhosphorImager (Personal Molecular Imager FX, Bio-Rad, Hercules, CA) to obtain high resolution images, which were then quantified with the Molecular Dynamics Storm & Array Vision Software (Incyte Genomics, Inc.).

Radioactive intensity measurements from each array element were compared with the median of local background, and the elements with values less than this median were excluded. Furthermore, for statistical analysis, we excluded elements with a correlation rate outside of 0.4 to 2.5. After these filters were applied, a total of 1,064 spots were evaluated. The expression ratios for the duplicated spots on the array were averaged. Furthermore, the ratios between atrophic endometrium and the other samples were calculated. We selected genes with expression levels that differed by a factor of at least 0.75- to 1.25-fold with respect to the reference sample (the atrophic pool). The significance for expression differences was computed by the F test of the two-tail test (P < 0.05), and after this analysis, we obtained a total of 53 genes. This group of selected genes showed a significant variation in expression levels across the patient samples and ensured that the genes considered showed relevant differences with respect to the pool. Therefore, the genes considered can be regarded as effectively repressed or overexpressed.

The set of differentially expressed genes was clustered by the similarity of their expression profiles. The Gene-cluster program was used for hierarchical clustering analysis (http://microarray.nhgri.nih.gov/genecluster).⁷ Using the Pearson’s correlation rate, we hierarchically arranged genes in function of their expression levels. The cluster was color coded, with red for up-regulation from atrophic endometrium and green for down-regulation.

Quantitative real-time PCR (TaqMan, Applied Biosystems, Foster City, CA) technology was used, to validate the differential expression of selected genes. Analysis was performed with the ABI PRISM 7700 Sequence Detection System Instrument and software (Applied Biosystems), with the manufacturer’s conditions [52°, 2 minutes; 95°, 10 minutes (95°, 15 seconds; 60°, 1 minute, for 50 cycles)]. The sequence of oligonucleotides and the TaqMan probe used for the analysis of RUXN1/AML1 were obtained with the Assays-on-Demand Applied Biosystems database (ref. Hs00231079 m1). The endogenous control, eukaryotic 18S rRNA (ref. C6 4308329; Applied Biosystems), was used to normalize variations in cDNA quantities from different samples. Each reaction was performed in triplicate with cDNA dilution from the atrophic pool (16 samples), cDNA dilution from the proliferative, healthy pool as calibration samples, and also, 30 cDNA dilutions of EEC. SDs, in triplicate, up to 0.380 were not accepted. All of the reactions were repeated at least twice. The quantity of target and endogenous reference was determined from a standard curve for each experimental sample. The standard curve was constructed, in each of the experimental repetitions, by 5-fold serial dilutions of cDNA 1, 1:10, 1:50, 1:100, 1:500, and 1:1,000. Standard curves not adjusted to 3.32 (± 0.1) were not accepted. The comparative standard curve and threshold cycle (C_t) method was used to calculate the amplification factor.

Nonaffected endometrial epithelium (atrophic endometrium) and tumor samples (EEC) from the same patients coded 13, 14, 15 in Table 2, were lysed and the supernatant boiled in Laemmli buffer. Samples were resolved on 12% SDS-PAGE gels, transferred onto a Inmobilon-P membrane (Millipore, Bedford, MA), and incubated overnight at 4°C with primary antibodies (1:1000 pAb α-RUNX1/AML1, ab11903, Abcam Ltd., Cambridge, United Kingdom; 1:400 mAb α-actin, Lab Vision-Neomarkers, Fremont, CA). Proteins were revealed with horseradish peroxidase-conjugated antibodies (DAKO, Glostrup, Denmark) and the SuperSignal West Dura substrate (Pierce Biotechnology Inc., Rockford, IL).

Tissue microarray immunohistochemistry was used to confirm the differential protein expression of RUNX1/AML1. We constructed three different tissue microarrays, namely atrophic, simple and complex hyperplasia, and EEC. Paraffin-embedded tissue samples were obtained from the Departments of Pathology from two institutions, Hospital Vall d’Hebron and Hospital del Mar. Representative areas from 74 endometrioid cancers (including 3 samples of stage IA, 29 of stage IB, 20 of stage IC, 7 of stage IIA, 6 of stage IIB, 6 of stage IIIA, and 1 of stage IIIB), 30 complex hyperplasias, 30 simple hyperplasias, and 15 atrophic endometria were carefully selected. Samples were hematoxylin-and-eosin-stained in sections and were marked on individual paraffin blocks. Two tissue cores of 1 mm in diameter were obtained from each paraffin block. The tissue cores were precisely arrayed in a new paraffin block, with a tissue microarray workstation MTA-1 (Beecher Instruments, Silver Spring, MD). Sections of 5 μm were obtained from all tissue microarray paraffin blocks.

RUNX1/AML1 was detected by indirect immunoperoxidase assay in the 5-μm paraffin-embedded sections, treated with heat in citrate buffer for antigen retrieval. Endogenous peroxidase activity was quenched with 3% H₂O₂; sections were incubated with the primary antibody against the NH₂-terminal RUNX1/AML1 (Santa Cruz Biotechnology, Santa Cruz, CA) for 1 hour at a 1:50 dilution and were washed and were thereafter incubated with peroxidase-conjugated rabbit antigoat immunoglobulin (EnVision, DAKO). Subsequently, sections were washed, and reactions were developed with diaminobenzidine, followed by counterstaining with hematoxylin. Quantitative and qualitative evaluation for RUNX1/AML1-positive immunostaining was performed by three independent investigators (J.P., M.D., and F.A.) by counting the number of positive cells and the number of total cells per a 1-mm core diameter (×400). Whenever a difference between the three assessments occurred, the slides were reviewed jointly, and a consensus was reached. The quantity of positive immunostaining was evaluated as follows: 0 = negative, 1 = low, 2 = medium, and 3 = high.

Statistical comparisons were performed with the Statistical Package for Social Science 11.5 (SPSS, Inc., Chicago, IL).

The significance of the different expression values was computed by the F test of the two-tail test (P < 0.05).

The expression level values in the pathological samples were distant from a normal distribution; therefore, different comparisons were analyzed with the nonparametric statistics U from Mann–Whitney. The significant value level was adjusted for multiple comparisons analysis with Bonferroni’s Method to reach the type I error, ≤0.05.

Comparisons between different groups of samples were analyzed with the nonparametric statistics U from Mann-Whitney. The significant value level was adjusted for multiple comparisons analysis with Bonferroni’s Method to reach the type I error, ≤0.05.

During fertile life, the endometrium is a permanently changing tissue, because of hormone level variations that occur in female physiology. It is known that the levels of ovarian steroids, as well as the levels of their receptors, can influence gene expression in the endometrium. Nevertheless, endometrial carcinoma presents mainly after menopause has occurred. It is the result of the transformation of normal epithelial cells following an evolutionary pathway, which includes, first, simple hyperplasia and, then, complex hyperplasia, before the cells finally reach the definitive characteristics of cancer. Previous analyses with microarray technology have focused on characterizing gene expression into different EEC histiotypes (20), but they have never studied expression profiles in simple and complex hyperplasias and in atrophic endometrium. As mentioned earlier, the molecular events that occur during the development of endometrial carcinoma, including microsatellite instability, DNA-altered methylation, as well as PTEN, K-Ras, and β-catenin status, are numerous, but none of them fully explains the cell transformation that takes place (6). To identify new molecular markers that may be responsible for the cell transformation of normal endometrial tissue into simple and complex hyperplasias and carcinoma, we have used microarray technology to study gene expression profiles in the above-mentioned specimens.

We used cDNA arrays containing 8,345 probes, most of which corresponded to known genes. Comparison of gene expression levels was made by hybridizing cDNA from four EECs and two pools of normal atrophic endometrium and normal proliferative endometrium (see Materials and Methods and Table 1). These two pools were designed to obtain a mean for gene expression levels, in both nontumoral postmenopausal and nontumoral premenopausal conditions, respectively. To avoid differential gene expression due to hormonal status, samples of normal, late-proliferative endometrium were selected from the same day of the menstrual cycle. To eliminate false positives, genes having expression levels as high as 1.25 or as low as 0.75, when normal proliferative endometrium was compared with atrophic endometrium, were discarded, because they were considered to be involved in the normal menstrual proliferative process. Differential gene expression values, ranging up to 1.25 or down to 0.75, were selected from the ratio between each analyzed EEC with respect to atrophic endometrium, and these were assumed as basal levels. After array processing and scanning, 53 cDNA sequences were found to be significantly altered (P < 0.05), corresponding to 0.635% of the total array (Fig. 1).

The hierarchical sample cluster reflected differences between histiotypes, showing three major branches. One represented normal proliferative endometrium (B/A; Fig. 1, orange in the upper hierarchical tree). Stage I EEC (C/A, D/A, E/A; Fig. 1, blue in the upper hierarchical tree) comprised a second branch, and a third one corresponded to stage II EEC (F/A; Fig. 1, black in the upper hierarchical tree). This analysis indicates that the gene expression patterns between EEC samples were somewhat variable, as expected, although differences were statistically significant enough to allow them to cluster into their respective histiotypes. This suggests that the different gene expression profile of each histiotype may reflect the evolution of the tumoral staging and the process of tumorigenesis in endometrial cancer.

On the other hand, the hierarchical gene cluster also showed two main branches, reflecting both the down- (green branch in the left hierarchical tree in Fig. 1) and the up-regulated (red branch in the left hierarchical tree in Fig. 1) genes in the EECs compared with atrophic endometrium. A third branch was also observable, including the genes having variable expression levels among the EEC samples. Furthermore, the two main groups of clustered genes showed two subgroups, depending on the expression levels among the high- and low-stage histiotypes. These subgroups included the down- or up-regulated genes in all of the analyzed histiotypes (genes 1–7 and 51–53 in Fig. 1), and also the down- or up-regulated genes in stage I EECs with respect to stage II EECs, in addition to atrophic endometrium (genes 8–29 and 37–50 in Fig. 1). The gene cluster analysis allowed us to identify a subset of 53 genes, significantly altered in EEC. Among these, 47 cDNA sequences were identified as known genes, and were classified into seven functional categories: 9% were involved in cell cycle regulation and cellular proliferation (i.e., DNAJA2, RPS4X), 17% were involved in transcriptional regulation (i.e., SNAPC1, TCEB3). 21% were involved in signaling (i.e., ADD3, NPTX1), 6% were involved in traffic (i.e., NCALD), 8% in the immune response (i.e., SIAT1), 9% corresponded to membrane proteins (i.e., MS4A6), and finally, 19% of the genes were involved in the cellular metabolism (i.e., BTN2A1).

Among the most up-regulated of the genes, RUNX1/AML1 was of special interest, because it is an oncogene, largely described to be related to acute myeloid leukemia. The RUNX1/AML1 gene (runt-related transcription factor 1/acute myeloid leukemia 1), a chromosome 21q22.3 hematopoietic regulator, is frequently translocated in leukemia (21). Its main translocations include AML1-ETO in t[8;21], CBFβ-SMMHC in inv[16], AML1-Evi-1 in t[3;21], and TEL-AML1 in t[12;21] (22). RUNX1/AML1 plays an important role in regulating the expression of many genes involved in hematopoietic cell development, and the impairment of RUNX1/AML1 function deregulates the pathways leading to cellular proliferation and differentiation (23). A stimulatory role of RUNX1/AML1 in G₁-S phase transition has already been demonstrated by Bernardin and Friedman (24). In their study, the stimulation of G₁ also induced apoptosis in 5 to 10% of the cells examined, an effect consistent with excessive G₁ stimulation (24). These authors also indicated that RUNX1/AML1 activates the transcription of one or more genes, critical to the G₁-S phase transition via its COOH-terminal transactivational domain. Inactivation of RUNX1/AML1 in acute leukemia is expected to slow proliferation, unless additional genetic alterations coexist, which accelerate G₁ (24). Point mutations of the RUNX1/AML1 gene present in the region, encoding the Runt domain, cause the loss of DNA-binding ability (25). Moreover, overexpression of the RUNX1/AML1 proto-oncoprotein in NIH3T3 cells has been shown to induce neoplastic transformation, depending on the DNA-binding and transactivational potencies (26). From the RUNX1/AML1 gene, two representative forms of proteins, a and b, are generated by alternative splicing. Both forms have the runt homology domain, possessing the DNA-binding ability, but only the b form causes neoplastic transformation (26). It is not known whether this occurs in endometrial carcinogenesis. Additional experiments are required to assess such a hypothesis. Finally, Ramaswamy et al. (27), characterized RUNX1/AML1 as one of the 17 genes associated with a molecular signature for metastasis in adenocarcinomas of different origins, including cancer of the uterus, the breast, and the ovaries. Even though endometrial cancer is easily detectable and late stages are becoming more and more rare, it is still unknown whether RUNX1/AML1 expression participates in endometrial metastatic dissemination.

To validate the cDNA microarray data, we examined the expression levels of the RUNX1/AML1 gene in a panel of 30 EECs, with TaqMan real-time quantitative PCR. Moreover, this panel was composed of tumors, ranging from low infiltrating stage IA to highly invasive stage IIIA (see Materials and Methods and Table 2). This allowed us not only to validate the up-regulation of RUNX1/AML1 in EEC, but also to characterize the distribution of gene expression levels, depending on tumor stage. RUNX1/AML1 expression was normalized against atrophic endometrium and normal proliferative endometrium calibration samples (see Materials and Methods).

As shown in Fig. 2, RUNX1/AML1 was clearly up-regulated in the EECs analyzed, both when normalized against atrophic endometrium (Fig. 2,A) and when normalized against normal proliferative endometrium (Fig. 2,B). Furthermore, when one assumes that the mean and the SD values obtained from the calibration samples define these populations better and that they should have fit into a normal distribution, with those values distant from the mean ± 1.96 SD considered atypical, RUNX1/AML1 was also found to be significantly up-regulated (Fig. 2 C).

Interestingly, among all of the up-regulated samples, the higher expression levels corresponded to the IC stage, which incorporates tumors with >50% of myometrial infiltration. Values from this tumor stage were found to be significantly elevated, when compared with noninfiltrating IA or to <50% infiltrating IB stages (P < 0.002). Moreover, RUNX1/AML1 up-regulation was also detected to a lesser degree in samples with stages II and III (P < 0.001), corresponding to tumors affecting the cervix, but not necessarily being infiltrating myometrium. Finally, the fact that the real-time quantitative PCR expression profile was similar, both when normalized against atrophic endometrium and when normalized against normal proliferative endometrium, reflects that the up-regulation found in RUNX1/AML1 was not dependent on the proliferative status of the endometrium. This validates the conditions imposed in the cDNA microarray analysis to specifically select for genes up-regulated in EEC and confirms the up-regulation of this gene.

RNA levels, demonstrating the up-regulation of RUNX1/AML1 in EEC, correlated with protein levels as shown by Western blotting performed with representative samples from the panel of EECs used for the real-time quantitative PCR characterization (Fig. 2,D). Tumor area from the selected samples corresponding to stage IC EEC (samples coded 13, 14, 15; see Table 2) was compared with its respective atrophic endometrial area obtained from the same patient (see Materials and Methods). Increased RUNX1/AML1 protein levels were reproducibly found in the different tumor samples analyzed when compared with their corresponding normal atrophic endometrium epithelia (Fig. 2 D).

To further characterize RUNX1/AML1 expression at the protein level, we prepared tissue microarrays, including atrophic endometrium, simple and complex hyperplasias, as well as endometrioid endometrial carcinoma (EEC) specimens. The tissue arrays allowed us to characterize RUNX1/AML1 protein levels in a wide variety of samples, representative of the most relevant tissue contexts found in postmenopausal patients with endometrial pathologies.

In normal atrophic endometrium, the nuclei of epithelial endometrial cells were found negative for RUNX1/AML1 labeling. Staining of the cytoplasm in the epithelium of atrophic-cystic glands was also negative but was focally and faintly positive in some multilayered glands, mainly in the perinuclear and apical areas. Surface epithelium showed a similar pattern (see further ahead on Fig. 4). Concerning EEC, an example of 35 samples, spotted in a tissue array and labeled for hematoxylin and eosin, is illustrated in Fig. 3. A 4× and 10× magnification of one of these spots, immunostained for RUNX1/AML1 and corresponding to a representative EEC, showed RUNX1/AML1 labeling in the epithelial glands and completely negative staining in the stromal compartment. The negative staining of RUNX1/AML1 in the stromal compartment increases the significance of the RUNX1/AML1 results obtained, because they are specific to the epithelial cells and obviate a possible stromal contribution within the different samples analyzed in the original cDNA micro-array screenings. A 60× magnification showed a major cytoplasmic staining of RUNX1/AML1, with additional punctuate, nuclear staining in some cells (Fig. 3). The striking cytoplasmic labeling found for this transcription factor might suggest the existence of a pool of the protein that eventually translocates to the nucleus, in which it is expected to exert its function. The specificity of this staining was verified by green fluorescence protein-RUNX1/AML1 transfection on an EEC cell line and immunocolocalization by confocal microscopy (data not shown). On the other hand, the nuclear labeling could reinforce the possible role of the up-regulated RUNX1/AML1 as a transcription factor in endometrial carcinoma. RUNX1/AML1 belongs to the transcription factor protein family characterized by the Runt domain, and is responsible for both DNA binding and heterodimerization with its partner protein, PEBP2β/CBFβ (28, 29). The function of RUNX1 as a transcription factor differs, depending on its interaction with different types of cofactors. On one hand, RUNX1/AML1 functions as a transcriptional activator of hematopoiesis- related genes, by interacting with the histone acetyltransferase coactivator, p300/CBP (30). On the other hand, RUNX1/AML1 can also behave as a transcriptional repressor, through its interaction with a corepressor (31). Harrington et al. (32) conducted a study in which RUNX1/AML1 was localized within the nucleus and was found to punctuate foci that are involved in transcriptional control and which associate with the subnuclear scaffold, designated as the nuclear matrix. The dynamic shuttling of RUNX1/AML1 within living cells to these stabilized foci has been proposed as a component of the mechanism for gene regulation in vivo (32). Our results suggest that the shuttling of RUNX1/AML1 between nuclear and cytoplasmic localizations could represent an additional level of regulation in EEC.

In the tissue array, when we compared atrophic endometrium with simple hyperplasia, complex hyperplasia, and EEC cores, we observed a significant gradation in the intensity of RUNX1/AML1 staining. Atrophic endometrium was mainly associated with negative staining, whereas in simple hyperplasia, we could find cores with a RUNX1/AML1 staining, ranging from intensity level 0 to 2 (Fig. 4). No negative staining was found, either in complex hyperplasia or in EEC, and only positive values, ranging in intensity level from 1 to 3 (Fig. 4), were observed. Moreover, when we quantitatively analyzed the intensity of the labeling, we could find significant differences between EEC samples and complex hyperplasia, simple hyperplasia, and atrophic endometrium samples (P < 0.001). In EEC, 30% of the samples corresponded to high-level intensity (intensity level 3), whereas only 5% was found to correspond thus in complex hyperplasia samples, and 0% in simple hyperplasia and atrophic endometrium samples (Fig. 5). A medium intensity level (intensity level 2) was also found, mainly in the EEC samples (26%), and was found at a lower percentage in complex hyperplasia and simple hyperplasia samples (Fig. 5). A low intensity level (intensity level 1) was mainly associated with hyperplasias, with 85% found in complex hyperplasia samples and 90% in simple hyperplasia samples (Fig. 5). Finally, negative intensity levels (intensity level 0) were practically restricted to atrophic endometrium samples (93%; Fig. 5). Similar results were obtained with antibodies raised against both the NH₂-terminal and the COOH-terminal part of the protein (data not shown).

These results not only confirm the overexpression of RUNX1/AML1 in endometrial carcinoma but also describe another level of gradation, already found among tumor stages (see Fig. 2). This gradation corresponds to an increase in RUNX1/AML1 up-regulation as the process of tumorigenesis progresses from normal atrophic endometrium to simple and complex hyperplasias and, then, on to carcinoma. As a DNA-binding transcription factor, RUNX1/AML1 could be regulating the expression of genes involved in EEC. RUNX family members have been implicated in transcriptional activation by acting as organizing factors at the promoters and enhancers of target genes where they associate with cofactors and other DNA-binding transcription factors that are required for gene regulation (33). Alternatively, RUNX proteins are potent repressors of transcription in a cell type-specific manner, resulting in either temporal transcriptional repression or irreversible epigenetic silencing (34). In this context, RUNX1 has been described to form stable complexes with corepressors as histone deacetylases and histone methyltransferases (35). Moreover, the kinetics of RUNX1/AML1 expression described in the real-time quantitative PCR sample panel, showing a peak at early myometrial invasion stages, led us to speculate that up-regulation of RUNX1/AML1 in EEC might be associated with promoter methylation as a mechanism of transcriptional silencing of genes involved in the first steps of tumor invasion of adjacent tissue. Finally, RUNX1/AML1 gene translocations are not excluded from being involved in EEC.

In conclusion, the data presented in this work clearly demonstrate the overexpression of RUNX1/AML1 and suggest that this gene, classically associated with acute myeloid leukemia, might play a particular function in the initial steps of myometrial infiltration in EEC. Transcriptional regulation of the human osteopontin promoter has been described by C/EBPα and AML1 in metastatic melanoma cancer cells (36). Moreover, the involvement of CBF/AML factors in CD11a/CD18 leukocyte integrin expression was proposed to have been deregulated in acute myeloid and B-lineage, acute lymphoblastic leukemias, thus contributing to their altered adhesion and metastatic potential (37). Concerning the down-regulation of RUNX1/AML1 present in the molecular signature of metastasis in primary solid tumors (27), we observed that the overexpression of this gene in EEC was specifically associated with the initial steps of myometrial infiltration. The decline in RUNX1/AML1 expression levels, found in samples corresponding to metastatic stages (Fig. 2), could hold information, regarding the principal role of RUNX1/AML1 in the promotion of metastasis. We are currently analyzing the possible role of RUNX1/AML1 in the initiation of tumoral cell invasion and in the promotion of metastasis in endometrial carcinoma.

The strategy presented here, in addition to its contribution to the understanding of the molecular events associated with endometrial tumorigenesis, provides more information on new target genes that could be useful in the diagnosis of this cancer. The fact that this gene could be specifically associated with the initial steps of myometrial infiltration, suggests that it may play a possible role as a new marker for prognosis and also as a new therapeutic target to improve EEC outcome. Nevertheless, additional experiments must be conducted, to unravel the mechanism of action for RUNX1/AML1 in endometrioid endometrial carcinoma.

The authors would also like to thank Drs. J. Vila (Servei d’Estadística, Institut Municipal d’Investigacions Mèdiques) and Ll. Armadans (Servei de Medicina Preventiva, Hospital Vall d’Hebron) for their help with the statistical analysis of the data. Also, additional thanks go to P. Gallego and M. Valeri (Unitat d’Alta Tecnologia, Institut de Recerca Vall d’Hebrón), and P. García (Servei d’Anatomia Patològica, Hospital del Mar) for their technical assistance, and to Lisa R. Piccione for her correction of the manuscript.