Main

Many different processes regulate cellular differentiation, proliferation arrest and, eventually, cell death. Neoplastic transformation is the result of deregulation of any of these processes that lead to abnormal proliferation of the cells. Chromatin structure and modifications and, therefore, chromatin regulatory factors, are involved in the control of cell growth and, consequently, in tumorigenesis. Among these factors, the polycomb group (PcG) constitutes an important set of proteins that regulate gene activity at the chromatin level.1, 2, 3 In Drosophila, in which they were first described, two groups of proteins were found to be responsible for the maintenance of homeotic gene activity in the appropriate segments during fly development: PcG itself, and the trithorax group (trxG).

PcG proteins form DNA-binding protein complexes, and although the composition of these is variable, two biochemical and major functional PcG complexes have been described in humans.

  • The Polycomb Repressive Complex 1 (PRC1), or PcG-maintenance (PcGm), which organizes the chromatin into a repressive structure.4, 5, 6, 7, 8 MEL18, BMI1, RING1, RNF2, HPH1, HPH2, HPC1, HPC2 and HPC3 are members of this complex. Recent studies identify a PRC1-like complex as the H2A ubiquitin ligase, linking H2A ubiquitination to PcG silencing.9

  • The PRC2 and PRC2-like complexes, which have histone methyl transferase activity, also known as PcG-initiating (PcGi), contain the proteins ENX/EZH, EED and SUZ12, among others.4, 10

Although the two complexes seem to have opposite roles with respect to cellular survival and proliferation, they are thought to function concertedly to maintain the repressed status of the gene. First, PRC2 methylates K27 and/or K9 of histone H3 and PRC1 complex recognizes this marker11 binding to chromatin.

Many different PcG complexes have been characterized that also contain non-PcG proteins such as transcriptions factors,12, 13 histone methyltransferases,11 and histone deacetylases.14 The composition of these complexes has been postulated to depend on cellular type and context.10, 15, 16

RYBP was identified by its direct interaction with RING1, RNF2 and YY1 proteins.8 RYBP is thought to belong to the PcG family, based on data from genetic analysis that implies a PcG function for this protein (MV, unpublished observations).

Defects of several of these proteins in transgenic mice have been shown to produce skeletal and hematopoietic defects and other specific alterations. For example, mice transgenic for Bmi1, Mel18 and M33 genes show hematopoietic anomalies such as severe spleen and thymus hypoplasia,17, 18, 19 and Rnf2 deficiency results in an early lethal phenotype.20

Studies of PcG in human tissues show that this system plays important roles in several biological functions, such as regulation of cell differentiation,16, 18, 21, 22 hematopoietic and neural stem cell self-renewal23, 24, 25 and control of cellular proliferation.3, 17, 18, 21, 26, 27, 28, 29

The role of PcG in human cancer is mostly unknown, but some evidence indicates the involvement of this system in tumorigenesis. Recent studies have demonstrated the importance of certain proteins in several types of human tumor. This is the case for EZH2 (with histone methyl transferase activity), which is involved in progression in prostate cancer and in neoplastic transformation of breast epithelial cells,30, 31, 32 and murine Bmi1, which collaborates with c-Myc in transforming lymphoid cells.33, 34, 35 Human BMI1 has been found to be deregulated in mantle cell lymphomas,36, 37 in Hodgkin's and in diffuse large B-cell lymphomas.38, 39, 40, 41, 42 A high level of expression of BMI1 has also been described in medulloblastomas,43 colorectal carcinoma,44 and a misregulation of BMI1 was shown to be inversely correlated with INK4/ARF in NSCLC.45 However, the existing studies have largely focused on lymphomas and leukemias,26, 37, 38, 46, 47 and no exhaustive analysis has hitherto been made of tumors of different origin.

Our aim was to analyze PcGm protein expression in a broad spectrum of tumoral samples, and to compare it with the expression in their corresponding normal counterparts, with the ultimate aim of investigating the role of this PcGm complex in human tumorigenesis.

Materials and methods

Samples

In total, 154 nontumoral and 550 tumoral samples of human tissues were collected from the archives of the Pathology Department from the ‘Nuestra Señora del Prado’ (Talavera de la Reina-Toledo), ‘Complejo Hospitalario Universitario de Vigo' (CHUVI)’ (Vigo-Pontevedra) and ‘Virgen de la Salud’ (Toledo) Hospitals, and from the tissue archives of the CNIO Tumour Bank Network. Paraffin blocks were selected on the basis of the availability of suitable formalin-fixed paraffin-embedded tissue (at least 1-mm thick). This study was carried out on anonymized archival biopsy material. The procedure complied with the guidelines of the Helsinki Declaration and the study was approved by the Ethics Committee of the Hospital Carlos III (National Center for Clinical Research).

Three different samples of each of the following nontumoral human samples were collected:

  • Central nervous system: parietal lobes, cerebellum, basal nuclei, brain stem, choroid plexus.

  • Gastrointestinal tract: esophagus, stomach, small bowel, colon, appendix, liver, pancreas, gallbladder.

  • Respiratory tract: parotid, larynx, trachea, lung (pleura and bronchi).

  • Endocrine system: pituitary, thyroid, parathyroid and adrenal glands.

  • Skin and soft tissue: skin, muscle.

  • Lymphoid tissue: tonsil, spleen, thymus.

  • Breast and gynecological tissue: breast, ovary, uterus (cervix, endometrium, myometrium), fallopian tubes, placenta, umbilical cord.

  • Male reproductive system and urinary tract: kidney, bladder, prostate, testis, seminal vesicles, epididymis.

  • Hyperplasic and inflammatory tissue: nodular thyroid hyperplasia, parathyroid hyperplasia, seborrheic keratosis, ulcerative colitis, hyperplasic polyp of the colon.

The tumoral samples included in the study represent the most frequent tumor types in humans, and include the following diagnoses:

  • Central nervous system: low- and high-grade astrocytomas, oligodendroglioma, ependymoma, meningioma (10 cases per type).

  • Gastrointestinal tract and annexes: gastric adenocarcinoma (10 well differentiated and eight poorly differentiated cases); gastric signet-ring adenocarcinoma (10 cases), colonic villous adenoma (10 low-grade and 10 high-grade dysplasia cases); colonic tubular adenoma (10 cases); well-differentiated colorectal adenocarcinoma (10 cases); mucinous colorectal adenocarcinoma (10 cases), poorly differentiated colorectal adenocarcinoma (5 cases); hepatocellular carcinoma (10 cases); gallbladder adenocarcinoma (9 cases); carcinoid tumor of the appendix (3 cases); endocrine pancreatic tumor (5 cases); exocrine pancreatic adenocarcinoma (5 cases).

  • Respiratory tract tumors: parotid pleomorphic adenoma (10 cases), squamous cell carcinoma of the larynx (10 cases), lung adenocarcinoma (10 cases), lung squamous cell carcinoma (10 cases), lung undifferentiated large-cell carcinoma (10 cases), lung neuroendocrine small-cell carcinoma (6 cases), mesothelioma (10 cases).

  • Endocrine tumors: pituitary adenoma (10 cases), thyroid papillary carcinoma (10 cases), thyroid follicular carcinoma (8 cases), thyroid medullary carcinoma (9 cases), adrenal cortex carcinoma (3 cases), adrenal pheochromocytoma (8 cases), parathyroid adenoma (10 cases).

  • Skin and soft tissue tumors: cutaneous basal cell carcinoma (10 cases), cutaneous squamous cell carcinoma (10 cases), malignant melanoma (10 cases), capillary haemangioma (10 cases), Kaposi's sarcoma (10 cases), fibrous solitary tumor (4 cases), leiomyosarcoma (6 cases), liposarcoma (7 cases) and atypical lipomas (3 cases).

  • Lymphomas: non-Hodgkin's lymphomas (follicular, mantle cell (MCL), diffuse large B-cell. Burkitt's (BL) and peripheral T-cell lymphomas) and Hodgkin's lymphomas (HL) (10 cases of each type).

  • Breast and gynecological tumors: breast ductal carcinoma (10 cases), breast lobular carcinoma (8 cases), squamous cell carcinoma of the cervix (10 cases), endometrial adenocarcinoma (8 cases) and serous endometrial carcinoma (2 cases), uterus leiomyoma (4 cases); ovary mucinous cystadenoma (5 cases) and cystadenocarcinomas (4 cases), ovary serous cystadenomas (4 cases) and cystadenocarcinomas (6 cases), ovary endometrioid adenocarcinoma (7 cases) and cystadenocarcinomas (3 cases), ovary clear-cell carcinoma (10 cases).

  • Tumors of the urinary and male reproductive system: bladder urothelial carcinoma, clear-cell renal-cell carcinoma (10 cases), prostate adenocarcinoma (10 cases), seminoma (10 cases), embryonal carcinoma (5 cases), teratocarcinoma (5 cases).

Tissue-Microarray Design

We used a Tissue Arrayer device (Beecher Instrument, Silver Spring, MD, USA) to construct the tissue microarrays (TMAs). All case samples were histologically reviewed and the most representative areas for each tissue or tumor type were marked in the paraffin blocks. Two selected 1-mm-diameter cylinders from two different areas were included in the TMAs. Thus, 10 different TMA blocks were constructed, two with normal tissue and eight with tumoral samples, each containing between 120 and 180 cylinders. Several controls were included in each block (placenta, testis, tonsil and tumoral cell lines) as internal controls to ensure the quality, reproducibility and homogenous staining of the slides.

The percentage of informative individual cores depends on the antibody evaluated. As each TMA included two different core cylinders for each case, the final percentage of missing expression data values was 10% for RING1, 5% for RNF2, 16% for BMI1, 11% for MEL18, 6% for RYBP and 3% for HPH1.

Immunohistochemical Techniques

The procedure has been described elsewhere.39, 42 Immunohistochemical staining was performed on these sections using the following antibodies (Abs):42 HPH1-α-mouse-monoclonal (Mc) Ab, RNF2 (3α)-mouse-McAb, RING1-rabbit-polyclonal (Pc) Ab, RYBP-rabbit-PcAb, BMI1 goat-PcAb (C20) (Santa Cruz Biotechnology Inc., CA, USA), MEL18 goat-PcAb (C20) (Santa Cruz) and for Ki67 the MIB1 antibody (DAKO, Glostrup, Denmark). Conditions for the demonstration of these proteins and controls for the antibodies have been described in other reports.7, 8, 29, 48, 42 Staining of TMA sections was evaluated by three different pathologists (ES, JG-C and MM), using uniform criteria. Discrepancies in the scoring of cases were resolved after joint examination on a multi-headed microscope. In order to guarantee the reproducibility of this method, we decided to employ straightforward, clearcut criteria, and cases were scored as positive (1 or 2) or negative (0) for every antibody (Table 1).

Table 1 Proteins and antibodies used in the study
Full size table

Extraction of Nuclei from Paraffin-Embedded Tissues

To study RING1 loss by FISH, we separated nuclei from paraffin cores, according to the protocol described by Paternoster et al.49 Briefly, we marked representative tumor areas on H&E-stained sections. Corresponding areas were identified on the paraffin blocks, and core tissue biopsies of 0.6-mm diameter were taken from each block using a tissue microarrayer needle (Beecher Instrument). The paraffin was dissolved at room temperature with three 20-min changes of xylene. The tissue was then rehydrated with 100, 95, 75 and 50% ethanol (EtOH) for 15 min each. The 50% EtOH was removed and the tissue was manually disaggregated with the tip of a paper clip. The tissue cores were digested by adding 100 μl of freshly prepared proteinase K solution (DAKO) and incubated at 37°C for 1–2 h depending on the tissue type. To aid enzymatic digestion, the sample was vortexed for 3 s several times during this incubation period. Nuclei were pelleted using a mini-microcentrifuge (7000 r.p.m.) for 10 min. Proteinase K was carefully removed with a micropipette and the nuclei washed in phosphate-buffered saline (PBS). The PBS solution was removed and the nuclei fixed by resuspension with vortexing of freshly prepared fixative (three parts methanol and one part glacial acetic acid). The nuclei were resuspended in 50 μl of fixative. Fixed nuclei suspensions were stored at −70°C until it was convenient to perform FISH studies.

Cytospins on glass slides were prepared by centrifugation of 25-μl cell suspensions+100 μl glacial acetic acid 70% (10 min, 1000 r.p.m.) in a Shandon Cytospin3 cytocentrifuge (Thermo Electron Corporation). The cytospin preparations were air-dried for 2–24 h. The samples were immediately processed for FISH, or stored in sealed boxes at −20°C before performing FISH.

Fluorescence In Situ Hybridization (FISH)

FISH analysis was carried out for the detection of BMI1 and RING1 copy number changes as previously reported for HER2 and c-MYC.50 In short, we used 4-mm sections of the TMAs for BMI1 and the cytospin described above for RING1 analysis. For the study of BMI1 amplification, we used the Bacterial Artificial Chromosome (BAC) clone RP11-573G6, from Research Genetics, Invitrogen Corp. (Carlsbad, CA, USA), which spans the entire 10p12.2 genomic region, together with a commercial centromeric probe for chromosome 10 (Vysis, Inc., DownersGrove, IL; USA) that was used as a control for the ploidy level for chromosome 10. For the study of RING1 losses, we used the BAC clone RP11-731H07, from Research Genetics, Invitrogen Corp., which spans the entire 6p21.3 genomic region, together with a commercial centromeric probe for chromosome 6 (Vysis, Inc).

Slides were deparaffinized, boiled in a pressure cooker with 1 mM EDTA, pH 8.0 for 10 min and incubated with pepsin at 37°C for 30 min. The slides were then dehydrated. Probes were denatured at 75°C for 2 min after overnight hybridization at 37°C in a humid chamber. Slides were washed with 0.4 × SSC and 0.3% NP40.

FISH evaluation was performed by two investigators (MCM and JCC) with no previous knowledge of other genetic, clinical or immunohistochemical results. Fluorescence signals were scored in each sample by counting the number of single-copy gene and centromeric signals in an average of 130 (60–210) well-defined nuclei. Amplification was defined as the presence (in >5% of tumor cells) of either >10 gene signals or more than three times as many gene signals as centromere signals of the chromosome. Deletion was defined as the absence of one signal of the gene compared with the centromeric probe signal from the same chromosome. Cutoff values for the copy-number changes were obtained from the analysis of normal adjacent cells in each experiment.

Results

PcG Protein Expression in Normal Tissues

The first objective was to investigate the distribution of PcG proteins in normal human tissues to establish the expression pattern of PcG in nontumoral tissues. To this end, we studied the expression of several PcG members of the PcGm complex (RING1, RNF2, BMI1, MEL18, HPH1 and RYBP) in 154 samples of nontumoral adult human tissues grouped in two TMAs.

The results revealed great variability in the distribution of most of the proteins in the tissues, and a very selective expression of some of these PcG components. Some tissues, such as testis, placenta and thyroid gland, expressed most of the proteins, while others, such as lung, liver and striated muscle, seemed to express only one or two proteins (Table 2 and Supplementary Table SI.1).

Table 2 Summary of PcG expression in normal tissues (For more information see Supplementary Table SI.1)
Full size table

The intensity and cellular-type distribution were different for each protein. In placenta, for example, the cytotrophoblast cells expressed MEL18, RNF2, BMI1 and RING1 and syncytiotrophoblast cells expressed RING1 and RYBP. HPH1 is the only PcG protein that was not detected in placenta (Table 2 and Supplementary Table SI.1).

Concerning the selective distribution of PcG proteins, some of them, such as RING1, were widely expressed in almost every tissue and cell type, while others were detected exclusively in a few specific cell types. For instance, RYBP was found in placenta, umbilical cord, thyroid gland and scattered cells in the pituitary gland, and HPH1 could be found in Langerhans islets, pituitary and parathyroid gland and germ cells in the testis. The results are summarized in Table 2 and Supplementary Table SI.1, and some examples are illustrated in Figure 1a, c, e and g.

Figure 1
figure 1

Examples of different patterns of PcG expression in normal and tumoral tissues: (a) pituitary gland; (b) pituitary adenoma; (c) testis; (d) embryonal carcinoma; (e) gastric surface-epithelial cells; (f) gastric adenocarcinoma; (g) kidney tubules; (h) clear-cell renal-cell carcinoma. (i and j) BMI1 amplification detected by FISH in an MCL (j) case overexpressing BMI1 (i). The picture shows a group of cells with several copies of BMI1 (stained in green) in comparison with two centromeric copies for chromosome 10 (red).

Full size image

Due to the proposed role of PcG in the control of cellular proliferation, Ki67, a marker of cellular proliferation, was also analyzed in these tissues. Levels of this marker were not clearly correlated with the presence of any PcG protein in tumors. Nevertheless, in normal lymphoid tissue, BMI1 was detected in quiescent cells (mantle cells) and MEL18 was present in the proliferating cells of the germinal center (GC) (Supplementary Table SI.1).

PcG Protein Expression in Human Tumors

The main goal of this study was to identify abnormalities in the expression of these proteins in human cancer, and, consequently, the role PcG protein deregulation may play in human tumorigenesis. Thus, we analyzed the expression of PcG proteins in the most frequent human tumors by immunohistochemistry to evaluate the anomalies in the expression of this system in tumors compared with normal tissues. We collected 550 tumoral samples (see ‘Samples’ in Material and Methods section) and made eight TMAs. A synopsis of the results is presented in Figure 2 and Supplementary Table SI.2.

Figure 2
figure 2

Expression of PcG proteins in human tumors (T) and in the corresponding normal tissue counterparts (N). Red color represents positive expression; green indicates the absence of expression. In the case of the tumors, the length of the red/green bars is proportional to the proportion of positive/negative cases, respectively, for each tumor type. Yellow indicates no samples.

Full size image

Changes in the expression of PcG between tumors and their normal-cell counterparts

Overall, tumor samples maintained the expression pattern of their normal counterparts. In agreement with this, RING1 protein was the most ubiquitously detected PcG protein in human tumors, and RYBP and HPH1 were undetectable in most tumors. However, some degree of heterogeneity was found in the staining for all proteins in specific tumors. The most significant observations made in this series are described below:

RING1: As previously mentioned, this protein was detected in every normal tissue type. When the tumors were analyzed, this ubiquitous expression was preserved with few notable exceptions: it was absent from only 3% of the tumors, the negative cases being of clear-cell renal-cell carcinoma (4/8 cases) (Figure 1h) and testicular germ-cell tumors (9/18), the latter including teratocarcinomas (1/4), embryonal carcinomas (2/4) (Figure 1d), and seminomas (6/10). These data are even more interesting given that RING1 was detected in the normal counterpart of the tumors, both in tubule cells in the kidney (Figure 1g) and in germ cells in the testis (Figure 1c).

RNF2: Analysis of the multitumor TMAs revealed some discrepancies between tumoral tissues and their normal counterparts. Such was the case for gastrointestinal tumors, whereby 77% of gastric (23/28) and 70% (40/55) of colonic tumors expressed RNF2, while it was not detected in the stomach or colon surface-epithelial cells (Figure 1e and f). GC-derived lymphomas (diffuse large B-cell lymphoma (DLBCL), Burkitt's lymphoma (BL) and Hodgkin's lymphomas (HL)) also had a higher level of RNF2 expression than that detected in B-GC cells.

BMI1: The expression of BMI1 was also variable in normal tissues and tumors, whereby 77% of tumors were positive for this protein. In gastrointestinal tumors, its expression compared with that of normal tissues was similar to that of RNF2. Accordingly, 63% of gastrointestinal tumors expressed BMI1, whereas it was only detected in sporadic colonic or gastric surface-epithelial cells (Figure 1e and f). BMI1 was also detected in several GC-derived lymphomas (DLBCL, BL and HL), whereas it was not expressed in nontumoral GC B-cells (Figure 2). MCL, pituitary adenomas and parathyroid adenomas were the tumors that expressed the highest levels of BMI1 protein compared to those detected in their normal cell counterparts (Supplementary Table SI.2).

BMI1 was not detected in 6/18 testicular germ-cell tumors (Figure 1d), and 4/9 renal-cell carcinomas (Figure 1h), whereas it was detected in germ cells in the testis (Figure 1c) and in tubule cells in the kidney (Figure 1g).

MEL18: This protein, together with RING1, was the most ubiquitously expressed in human tumors. Only 8% of the tumors were found to be negative for this protein. The tumoral types with a higher percentage of negative samples were cutaneous squamous-cell carcinoma (4/10), ovary clear-cell carcinoma (6/8), endometrial adenocarcinoma (6/8) and squamous-cell carcinoma of the cervix (4/10). MEL18 was detected in the corresponding normal counterparts (Figure 2).

RYBP: Only 10% of all tumor cases were found to be positive for RYBP, the cases corresponding to the following: oligodendroglioma (4/10), pituitary adenomas (9/9), Hodgkin (3/10) and T-cell lymphomas (3/9). RYBP was not detected in glial cells, or in normal lymphoid B- or T-cells.

HPH1: 10% of the tumoral cases were positive for HPH1 protein. In the central nervous system, HPH1 was detected in 6/9 low-grade compared with 2/10 high-grade astrocytoma cases, and in oligodendroglioma (2/10) and ependymoma (6/10). The normal glial cells did not express the protein. HPH1 was also detected in 9/10 pituitary adenomas, and in nearly every testicular germ-cell tumor (18/19) (Figure 1d). HPH1 protein was also detected in the pituitary gland and the germ cells of the testis.

FISH study

The level of BMI1 protein expression was found to be particularly high in three tumor types: MCL (6/10), pituitary adenoma (9/9) and parathyroid adenomas (8/10). Such a level of expression has previously been associated with genomic amplification in MCL.36, 37 Therefore, we decided to study this possibility in these three tumors types by FISH analysis in TMA paraffin sections. Our findings confirmed amplification of BMI1 locus in MCL in two of 10 cases, six of them with high BMI1 protein expression, and in one case out of nine pituitary adenomas. No amplification was detected in any of the 10 parathyroid adenomas analyzed, eight of which had high BMI1 levels.

As mentioned before, the loss of RING1 expression, only observed in 3% of all the tumors analyzed, was very selective for germinal-cell tumors and clear-cell renal-cell carcinomas. Therefore, we also employed FISH analysis to look for loss of RING1 locus in these tumors. To do this, we decided to separate nuclei from the whole paraffin sections of the tumors included in the TMAs as described in the Material and Methods section. In total, 16 cases were analyzed in this way: 13 negative cases and three positives as controls (clear-cell renal-cell carcinoma (five negative/two positive), seminomas (6/0), teratocarcinomas (1/1) and embryonal carcinomas (2/0)). Loss of RING1 locus was not detected in any case.

Unique PcG protein expression in pituitary adenomas, clear-cell renal-cell carcinomas and germ-cell tumors

There are some examples of tumors with a unique PcG expression pattern (Figures 1 and 2).

Testicular germ-cell tumors showed a distinctive PcG pattern with HPH1 expression at the highest levels and a parallel absence of RING1 protein and also of BMI1 (Figure 1d).

Renal-cell carcinoma was the other tumor type lacking RING1 expression, and also frequently showing absence of RNF2, BMI1 and MEL18. HPH1 was not detected in these tumors (Figure 1h).

Pituitary adenoma was the only type of tumor that coexpressed every PcG protein analyzed in this study (Figure 1b). Every valid case expressed all the proteins, except one that was negative for HPH1 (Supplementary Table SI.2).

Discussion

Our analysis of PcG protein distribution in human normal and tumoral tissue allows us to identify a PcG protein-expression profile in human normal tissues and tumors that implies a role for some PcG genes in specific neoplasms. Additionally, the anomalous high level of expression of BMI1 associated with gene amplification strongly supports a role for this PcG protein in cancer.

The distribution of several PcG proteins differed greatly in normal human tissues and among different cell types in specific tissues. HPH1 was detected exclusively in germ cells in testis, in pituitary and parathyroid gland and the pancreatic islets of Langerhans, and RYBP was found in placenta, umbilical cord, thyroid and pituitary glands. By contrast, RING1 was ubiquitously expressed in normal tissues, although some variability among different tissues and cell types was observed. MEL18, BMI1 and RNF2 had a more variable tissue distribution.

Overall, tumor samples shared the same expression pattern as their normal counterpart, although changes in the expression of PcG proteins associated with tumoral transformation were found for each protein at least in some tumors. There were several particularly remarkable findings associated with this tumoral transformation: (a) RING1-loss, which seemed to be a very selective finding in clear-cell renal-cell carcinoma and testicular germ-cell tumors: (b) MEL18 loss, which was characteristically found in several gynecological tumors, such as squamous-cell carcinoma of the cervix, uterine endometrial adenocarcinoma and ovarian clear-cell adenocarcinoma. This finding is consistent with the previously described loss of MEL18 in tumoral cell lines.51 (c) In contrast, other PcG proteins were overexpressed in a number of tumors. This was the case for RNF2 and BMI1, which were both detected in gastrointestinal tumors or GC-derived lymphomas, according to previously published data.37, 44 Although BMI1 was detected in mantle cells in reactive tonsils, its levels in MCL were also anomalously high, consistent with the amplification of this gene36, 37 confirmed in this study by FISH. High levels of BMI1 were also detected for the first time in pituitary and parathyroid adenomas, and gene amplification was confirmed in pituitary adenomas; and (d) HPH1 and RYBP were detected, with a mutually exclusive pattern, in tumors of the central nervous system. RYBP was also found in HL, as previously described,42 and in peripheral T-cell lymphomas.

The remarkable unique PcG expression profiles observed in particular tumors merit special attention. Testicular germ-cell tumors were one such type, with an absence of RING1 protein expression in 50% of cases, and a loss of BMI1 in 67% of cases. These are the tumors that expressed HPH1 most frequently, in consonance with the observations made in normal germ cells, expressing the highest levels of HPH1. Clear-cell renal-cell carcinoma is another tumor type lacking RING1 expression that frequently features an absence of RNF2, BMI1 and MEL18.

Pituitary adenomas had a very singular and distinctive PcG expression profile since they were the only tumor type expressing the six PcG proteins analyzed in almost every case. Additionally, strong expression of BMI1 was detected in all these cases, and was associated with gene amplification in one of them.

One of the points to be explored arising from these observations is that the composition of the PcG complex could vary among different cell types and could be regulated during tumoral transformation. This has been proposed in earlier studies16, 52 and our findings support the existence of tissue and cell-specific PcG complexes. Even the proteins grouped in the same PcG complex, as in the case of the proteins included in the present study, which belonged to the PcGm complex, could form different complexes depending on the cell type or differentiation status, and during tumor development (Figure 3).

Figure 3
figure 3

Proposed PcG components of the PcGm complex in several normal and tumoral tissues. The expression of various PcG proteins in different cell types allows us to propose the presence of different components of PcGm members in distinct normal and tumoral tissues.

Full size image

The selective distribution of the different components of the PcG complex in normal tissues and tumors could help to clarify their role as epigenetic chromatin modifiers, since changes in the methylation status, including those of nucleosomal histones, have been proposed as being universal in cancer cells.53

These data are evidence of the significant involvement that has been proposed for BMI1, by which it plays a role in permitting the self-renewal capacity of cancer stem cells.23, 43, 54 The relatively generalized increase in the expression of BMI1 in different tumor types compared with their normal counterpart cell types and the gene amplification detected in some tumors in the present study, sustains the interpretation that BMI1 occupies a key position in cancer development. These results provide no evidence in support of the proposed hypothetical role55 for BMI1 or MEL18 as cell-cycle regulators in tumors.