[page 97↓]

4.  Discussion

4.1. Yeast Two–Hybrid System

These studies were driven by the desire to understand the mechanism of action of the class II tumor suppressor gene H-REV107-1. Since the rat H-rev107 gene has been cloned, several reports were published about H-REV107-1 and its related genes in different species (Hajnal et al., 1994; Kuchinke et al., 1995; Sers et al., 1997; Husmann et al., 1998; Di Sepio et al., 1998; Akiyama et al., 1999; Huang et al., 2000; Ito et al., 2001; Sers et al., 2002). The main common features of these genes are that they are down-regulated in tumors and tumor derived cell lines, and possess growth suppressive properties. The primary sequence of the H-REV107-1 product did not provide any clues to the mechanism of its function as a growth suppressor. In order to elucidate the mechanism of human H-REV107-1 mediated growth inhibition, we performed a search for interacting proteins using a yeast two-hybrid assay.

Yeast two hybrid systems were developed in 1990by Stanley and his colleagues, and provided a novel way of detecting protein-protein interactions in vivo. One of the first examples of its application was the identification of the Raf Kinase as a Ras interactor and a downstream effector of Ras function (Vojtek et al., 2003).

We have used a LexA-based yeast two-hybrid system supposed to produce fewer false positives as compared to the GAL4-based system. This proposition is based on the fact that in the LexA system the DNA–binding domain (BD) and the activation domain (AD) are provided entirely by the prokaryotic LexA, and B42 E. coli proteins, respectively. The prokaryotic proteins are supposed to interact with very few of yeast or mammalian proteins over-expressed in yeast cells. In contrast, the GAL4 system contains plasmids with eukaryotic GAL4-DNA binding and activation domains which are supposed to be less selective. Nevertheless, some interactions can be recognised only by using a LexA-based yeast two hybrid system, and other interactions can be identified only with the help of the GAL4-based system (Golemis et al., 1994). The reason for this observation is unknown.

Prior to a large scale transformation for a library screening, the human H-REV107-1 protein was kindly tested by E. Cuppen (Department of Cell Biology, Institute of Cellular Signalling, University Nijmegen, The Netherlands) in an established yeast two hybrid approach. This test demonstrated that for a successful application of the approach, the truncated form of the H-REV107-1 protein without the membrane binding domain has to be used. Expression of the full length protein in yeast did not resulted in colony growth. We supposed that the membrane binding domain hampered a transport of the protein into the yeast nucleus where the interactions with library proteins take place. Therefore, we performed the entire screen using a truncated form of the H-REV107-1 protein, lacking the 27 C-terminal aminoacids.

[page 98↓]

The use of a truncated protein bears the risk of loosing some interactors, which bind to the deleted region. However, it was rather doubtful that a transmembrane domain will participate in protein-protein binding. Most investigations of protein-protein interactions demonstrated that membrane bound domains do not participate in interaction with cytoplasmic proteins. There are only a few examples demonstrating that such domains can be involved in interaction with other proteins, or in the formation of homodimers within a membrane (Langosch et al., 2002). The bacterial protein glycophorin is known to interact with other proteins via its transmembrane domain, although the mechanism of this protein-protein recognition is unclear (Gerber and Shai, 2002). In eukaryotes, large transmembrane proteins participating in transmembrane channel organisation were demonstrated to form homo- and heterodimers via the transmembrane regions (Nakayama et al, 2002).

Although the LexA-and GAL4-based yeast two hybrid systems are powerful tools for detecting protein-protein interactions, they have several disadvantages. They produce false positives, when library proteins intrinsically modulate transcription, for example when they function as transcriptional activators or repressors. In addition, the interaction between bait and pray takes place in the yeast nucleus, where proteins do not undergo posttranslational modifications like terminal glycosylation, sulfation, phosphorylation or methylation. Interactions that require protein posttranslational modifications in the cytoplasm can not be detected by these systems, as well as interactions which require modulating factors present in the cytoplasm or the cell membrane. These aspects were very important for our assay, because we supposed that the human H-REV107-1 protein is subject to posttranslational modification. Five potential tyrosine phosphorylation sites and one serine site were identified with a score more than 0,6 in the H-REV107-1 protein sequence (Fig. 31) using the NetPhos 2.0 program developed by the Technical University of Denmark (http://www.cbs.dtu.dk/services/NetPhos/). An additional search for potential myristoylation, glycosylation or tyrosine sulfation sites of the H-REV107-1 protein revealed negative results. Furthermore, a Western blot analysis of COS-7 cells transiently transfected with an H-REV107-1 expression vector revealed two bands of a slightly different mobility in SDS-PAGE (Fig. 12), which also suggests some kind of posttranslational modification.

Thus, the Lex-A yeast two hybrid assay identified only putative interactors of H-REV107-1 which can bind to the non-modified protein.

[page 99↓]

Fig. 36 Prediction of protein phosphorylation sites of the H-REV107-1 protein

Five tyrosine and one serine phosphorylation sites were identified using NetPhos program. These residues are shown in red.

Unfortunately, after we had established the LexA yeast two-hybrid system, several new two-hybrid approaches were developed, allowing an improved search for interacting proteins. One example is the CytoTrap two-hybrid system, which enables the discovery of protein-protein interactions in the cytoplasm of yeast cells. This system is based on the activation of Ras signaling pathways in a temperature-sensitive yeast cell line via the recruitment to the cellular membrane of human Sos protein (hSos) fused to the bait. The target proteins are anchored to the membrane. Therefore, interaction between bait and target leads to a close proximity of the hSos and Ras proteins. The hSos protein activates Ras by GDP/GTP exchange, and allows the yeast cells to grow at 37°C. The system was developed for the identification of interacting partners of the Jun protein (Aronheim et al., 1997). It is very well adapted for the identification of protein interactions which take place in the cytoplasm. An interesting modification of this system is, the so called “reverse Ras recruitment system”, in which the bait is anchored to the cellular membrane, and the proteinpartner (the prey) is fused to a cytoplasmic Ras mutant. Protein–proteininteraction between the prey and thebait results in Ras membrane translocation and activation ofa viability pathway in yeast (Hubsman et al., 2001).

One of the possibilities to identify interactions requiring tyrosine phosphorylation is the so called “TK (tyrosine kinase) two-hybrid system”. This system is very similar to the standard LexA-based approach. The difference is that an additional tyrosine kinase domain is ligated into the LexA-vector, containing the bait. Expression of this vector in yeast leads to the expression of a bait, and a tyrosine kinase which then phosphorylates the bait protein. Thus, using this system, a specific search of interacting partners of the tyrosine phosphorylated protein can be performed (Vojtek and Hollenberg, 1995).

[page 100↓]

We screened a human kidney cDNA library with a truncated form of the H-REV107-1 protein. After rescue of the yeast plasmids from positive colonies, and the following sequencing analysis, we searched for homology of the isolated cDNA inserts with known genes in the NCBI database to chose targets for further investigation. This step of analysis is the most critical in the two-hybrid assay. Among the isolated 168 clones, only a few sequences were present more than one time. We excluded typical predicted false positives from further analysis, although it might lead to the loss of putative true positives. For example, heat shock proteins were described to bind many targets unspecifically in yeast (Golemis et al., 1994). Therefore, we excluded the Hsp90 heat shock protein from further analysis. However in many publications it was demonstrated that Hsp90 is a molecular chaperone whose association is required for the stability and function of multiple mutated, and over-expressed signaling proteins that promote cancer growth and cell survival. Hsp90 client proteins include mutated p53, Bcr-Abl, Raf-1, c-Src, Akt, HER2 (Neckers, 2003). It participates in the prevention of tumor cells from apoptosis (Rashmi et al., 2003). Inhibition of Hsp90 results in induction of apoptosis through the activation of caspases–9 and -3 in AML HL-60 cells because of the attenuation of Hsp90-interacting proteins, including Akt, c-Raf-1, and c-Src protein kinases (Nimmanapalli et al., 2003). We demonstrated that ectopic expression and induction by IFNγ of H-REV107-1 in ovarian carcinoma cells leads to the induction of apoptosis through the activation of caspase-9 (Fig. 36). Therefore, in view of these findings, an interaction between Hsp90 and H-REV107-1 might also be plausible and should not entirely be excluded.

We also found a number of unknown genes, which we did not prove in a mating assay, and which were not further analysed in mammalian cells. Putative H-REV107-1 interacting partners might be among them. Therefore we recently repeated the BLAST search of sequences found three years ago. For most inserts still no homologies were found. However, the analysis revealed one novel gene, TSGA10 isolated in 2001 (Madarressi et al., 2001) without any known function. In addition two cDNA sequences encode proteins which were characterised in the meantime. One of them is the acid cluster protein ACP33, isolated as a novel intracellular binding partner of CD4, proposed to modulate a stimulatory activity of CD4 in T cells (Zeitlmann et al., 2001). The second cDNA insert encodes 100 amino acids at the C-terminus of the novel human protein Scribble, a homologue of the Drosophila protein scrib. Scrib dysfunction results in the false distribution of apical proteins (Bilder and Perrimon, 2001). It has been demonstrated that scrib acts as a tumor suppressor and participates together with two other tumor suppressors, lethal giant larvae (lgl) and discs-large (dlg) in the regulation of cell polarity and growth control (Bilder et al., 2000). The human homologue, hScrib, was shown to be a target of the papillomavirus (HPV) E6 protein, which stimulates its ubiquitination and degradation (Nakagawa and Huibregtse, 2000).

[page 101↓]

Based on the performed BLAST search analysis, and available literature about found potential interacting partners of H-REV107-1, we chose 11 clones for a mating test (Table 5). The mating assay is a powerful supply for the verification of false positives from the yeast two-hybrid system. Using this assay, we tested whether induction of the reporter genes is a specific response based on the interaction between H-REV107-1 and the corresponding target protein. The assay resulted in the exclusion of 5 proteins from further analysis, because their ability to activate reporter genes was independent of the H-REV107-1 expression (Fig. 7).

For further conformation in the mammalian system, we chose 6 proteins, namely PC4, PR65, RARG, S100A6, ETF1, and P14.5. The aim of the investigation was to prove if the interactions identified in yeast also occur in mammalian cells, and to define a role of the confirmed protein-protein bindings in the H-REV107-1 - mediated growth suppression, and induction of cell death. We verified that H-REV107-1 protein interacts with PC4, PR65, RARG, S100A6, and ETF1 proteins, but not with the P14.5 protein in transfected COS-7 cells.

4.2. H-REV107-1 is a Target of IRF-1 and Modulates IFNγ - Dependent Inhibition of Cellular Growth by Different Mechanisms

Recovery of H-REV107-1 expression upon IFNγ-treatment was demonstrated in human ovarian carcinoma cell lines. Further experiments revealed that H-REV107-1 is a target of the interferon regulatory factor 1, IRF-1 (Sers et al., 2002). Therefore we asked how H-REV107-1 is involved in the known IFNγ-dependent pathways leading to growth suppression and apoptosis.

Interferon gamma (IFNγ) is a cytokine which was originally identified as the protein responsible for the induction of cellular resistance to viral infection. Subsequently, much evidence has been accumulated with regard to its role in cell growth and differentiation (Pestka et al., 1987). Later, the IFNγ - response has been also postulated to be part of an endogenous tumor surveillance system (Coughlin et al., 1998). It exerts inhibitory effects on tumor cell growth, and recently an improved survival of ovarian carcinoma patients after therapy with IFNγ was described (Windbichler et al., 2000). The biological effect of IFNγ is mediated through a heterodimeric transmembrane receptor which activates a Janus kinase (JAK) – STAT pathway. JAK activates signal transducer and activator of transcription (STAT1) through tyrosine phosphorylation at the cell membrane, followed by dimer formation and migration of the STAT1 homodimers to the nucleus (Stark et al., 1998). Phosphorylated STAT1 enhances the recruitment of transcriptional coactivators, such as P300/CBP, to the promoters of the IFNγ target genes inducing their transcription (Paulson et al., 1999).

[page 102↓]

Important mediators of the IFNγ response are the STAT1 – target genes, encoding the interferon regulatory factors (IRFs). To date, nine members of the IRF family have been determined, IRF-1 – IRF-9 (Harada et al., 1998). IRF-1 and IRF-2 were identified originally as a transcriptional activator and its antagonistic repressor, respectively, mediating IFN-α, and -β signaling (Harada et al., 1989). Further investigations showed that IRF-1 and –2, both play a key role in cellular growth control, susceptibility to tumorigenic transformation, and induction of apoptosis. Consequently they were suggested to function as a tumor suppressor (IRF1) and oncogene (IRF2) (Harada et al., 1993; Sato et al., 2001).

The tumor suppressor activity of the IRF-1 gene is further supported by its localisation at chromosome 5q31.1, a region frequently deleted in human leukemias (Willman et al., 1993). The loss of one IRF-1 allele has also been reported in oesophageal and gastric cancer (Nozawa et al., 1998). Other possible mechanisms of IRF-1 inactivation may be alternative splicing of the IRF-1 mRNA, producing aberrant IRF-1 in human myelodysplasias and leukemias (Harada et al., 1994). In breast and ovarian carcinomas the IRF-1 gene exhibits features of a class II tumor suppressor. Down-regulation of IRF-1, similar to H-REV107-1, has been demonstrated in high grade human ductal carcinomas and in invasive breast cancers (Doherty et al., 2001; Sers et al., 2002). Significant reduction of the IRF-1, and H-REV107-1 mRNA level was revealed in the ovarian carcinoma cell lines OVCAR-3, A27/80, and PA-1 compared to the non-tumorigenic ovarian epithelial cells HOSE (Sers et al., 2002).

Abrogation of the anti-oncogenic IRF-1 activity can also be achieved by inhibiting of its DNA binding ability via direct interaction with a putative ribosome assembly factor, nucleophosmin (NPM)/B23/numatrin, over-expressed leukemias human leukaemia cell lines (Konde et al., 1997). Summarising these data, IRF-1 is a critical tumor suppressor gene, whose inactivation through various mechanisms contributes to the promotion of several human cancers.

The precise nature of the IRF-1 – dependent tumor suppression is not very clear, it is supposed that IRF-1 acts through the up-regulation of a set of genes whose products function as negative regulators of cellular growth (Harada et al., 1993). A number of IFN-stimulated genes which are involved in negative regulation of cell proliferation, have been shown to be IRF-1 targets. Among them are 2-5A synthetase, cyclin-dependent kinase inhibitor p21WAF1 (Coccia et al., 1999), lysyl oxidase (Tan et al., 1996), double-stranded RNA-dependent protein kinase, PKR (Beretta et al., 1996), and H-REV107-1 (Sers et al., 2002).

We demonstrated that H-REV107-1 is a direct target of IRF-1 in NIH3T3 cells harbouring estrogen-inducible IRF-1, and in a subset of ovarian cancer cell lines (Sers et al., 2002). Investigation of the H-REV107-1 and IRF-1 expression demonstrated their low level in A27/80 and OVCAR-3 cell lines, which was enhanced after IFNγ-exposition.

[page 103↓]

However, IFNγ suppresses growth of only A27/80 cells, OVCAR-3 cells were demonstrated to be resistant to the treatment (Sers et al., 2002). Analysis of the phenotype of the treated A27/80 and OVCAR-3 cells revealed no changes in A27/80 cells after 48 hours of incubation with IFNγ (data not shown). In contrast approximately 5% of the OVCAR-3 cells revealed a strong up-regulation of the H-REV107-1 expression, which correlated with an apoptotic morphology of nuclei (Fig. 15). This supposed that H-REV107-1 is directly involved in IFNγ - mediated apoptosis in OVCAR-3 cells. However, due to the low number of cells up-regulating H-REV107-1 after IFNγ - treatment this effect had remained undetected in cell growth assay performed earlier.

To investigate the mechanism of H-REV107-1 – mediated growth suppression, A27/80 and OVCAR-3 cells were transiently transfected with H-REV107-1 expression vector. In both cell lines apoptotic nuclear morphology was observed only in cells expressing H-REV107-1 protein (Shayesteh al., 1999; Sers et al., 2002). Thus, forced H-REV107-1 expression leads to a cell death in both cell lines, whereas IFNγ is likely to suppress growth of these cells by different mechanisms. We observed increase of IRF-1, IRF-2 and H-REV107-1 protein levels upon IFNγ - induction only in OVCAR-3 cells (). In contrast in A27/80 cells expression of these proteins was below detection limits (data not shown). To define further the IFNγ - dependent signaling in A27/80 and OVCAR-3 cells, we analysed the expression of the STAT1 protein, which, as was previously described, mediates most of the IFNγ-responses (Stark et al., 1998). In addition we analysed expression of the p21WAF1 cyclin dependent kinase, a central mediator of growth arrest and senescence in mammalian cells (Waldman et al., 1995). Increased p21WAF1 expression leads to cell growth arrest which occur in both G1 and G2 phases of cell cycle (Niculescu et al., 1998), and is accompanied by the development of morphologic and phenotypic markers of senescence (McConnell et al., 1998). p21WAF1 is regulated in response to DNA damage in a p53-dependent manner, but also via IRF-1 (Tanaka et al., 1996). The p53-independent induction p21WAF1 expression in response to IFNγ is mediated by STAT1, through direct binding of IRF-1 to the p21 promoter (Coccia et al., 1999). We have asked whether p21WAF1 participates in the IFNγ-response in OVCAR-3 and A27/80 ovarian carcinoma cells. We observed up-regulation of STAT-1 expression in both cell lines 24 hours after IFNγ treatment, and even enhanced level after 48 hours. Notably, up-regulation of p21WAF1 was observed only in A27/80 cells (Fig. 16). This result supported our hypothesis of different mechanisms of IFNγ-growth inhibition in OVCAR-3 and A27/80 cells.

We proposed the following model of IFNγ-response in OVCAR-3 cells: cytokine induction leads to the activation of STAT1, which stimulates IRF-1 expression. IRF-1 induces H-REV107-1, and IRF-2 transcription.

[page 104↓]

IRF-2 activates a negative feed-back loop, probably suppressing expression of H-REV107-1 and other IRF-1 targets. Expression of IRF-2 protein was detected already 24 hours after IFNγ treatment, suggesting rapid reverse of the IRF-1 effect (Fig. 14). It is still an open question why a minority of cells express high level of H-REV107-1 upon IFNγ-induction, and undergo apoptosis, whereas other cells are resistant to the IFNγ-treatment.

Although we observed up-regulation of mRNA H-REV107-1 and IRF-1 genes in A27/80 cells after IFNγ-induction, the amount of synthesised proteins was below the sensitivity of the method. Alternatively, the proteins were rapidly degraded, before we could perform Western blot analysis. The ubiquitin-proteasome pathway has been reported to play a key role in the down-regulation of the mouse IRF-1 protein (Nakagawa and Yokosawa, 2000). To prove if this pathway mediates degradation of the human H-REV107-1 and IRF-1 proteins, we treated several ovarian tumor cell lines with the MG115 and MG132 protease inhibitors after induction with IFNγ. We observed a stabilisation of the proteins only in human teratocarcinoma cells PA-1 but not in OVCAR-3 and A27/80 cell lines (data not shown). Thus, a protein destabilisation in PA-1 cell line is proteasome dependent, whereas in OVCAR-3 and A27/80 cell lines other mechanisms are responsible for the destabilisation of the H-REV107-1 and IRF-1 proteins.

Analysis of IRF-1, H-REV107-1, and STAT1 expression in A27/80 cells rather suggested an IRF-1 independent growth suppression in A27/80 cells after IFNγ - treatment. Observed up-regulation of p21WAF1 suggested that this protein might participate in the IFNγ-response. In addition to p53 and IRF1, the alternative regulator of the p21 WAF1 expression was demonstrated to be the breast cancer susceptibility gene 1, BRCA1 (Ouchi et al., 1998).

The tumor suppressor BRCA1 has been reported to be implicated in the DNA-repair process (Sculli et al., 1997), and in growth control specifically in breast and ovarian cancer cell lines, but not in colon and lung cancer cells or fibroblasts (Holt et al., 1996). Direct interaction of BRCA1 with the p300/CBP coactivator and with RNA polymerase II holoenzyme suggested that BRCA1 also plays a role in transcriptional regulation (Pao et al., 2000). Later findings have also implicated BRCA1 as a transcriptional regulator of the P21 WAF1 and MDM2 genes harbouring a p53-responsive element in their promoter regions. This suggested that BRCA1 can enhance p53-dependent gene regulation (Ouchi et al, 1998). Moreover, p21WAF1 was reported to be required for the BRCA1-mediated growth suppression (Somasundaram et al., 1997). It has been demonstrated that activation via IFNγ leads to the interaction between STAT1 and BRCA1 proteins, which stimulate the P21 WAF1 gene transcription independent of IRF-1 (Ouchi et al., 2001). Thus, BRCA1 is a critical component of the IFNγ - regulated anti-tumor response, and a possible regulator of the P21 WAF1 expression in A27/80 cells.

[page 105↓]

The status of the BRCA1 gene in OVCAR-3 and A27/80 cells is unknown, but regarding the failure of the p21WAF1 up-regulation in OVCAR-3 cells after IFNγ-induction, it is likely that BRCA1 is mutated in this cell line. In A27/80 cells BRCA1 might participate in p21 WAF1 induction.

Transcriptional activity of BRCA1 has been demonstrated to be maximal in the presence of PC4, although direct interaction between PC4 and BRCA1 was not shown (Haile and Parvin, 1999). We hypothesised that H-REV107-1 might serve as a PC4, and BRCA1, STAT1 binding protein. Such a prediction arose from the fact that we found the PC4 transcriptional coactivator in yeast two-hybrid system as a true H-REV107-1 interacting partner. PC4 serves as a potent coactivator of a diverse group of transcriptional activators in standard in vitro transcription systems (Ge et al., 1994). It interacts both with a variety of activation domains and with members of the RNA II polymerase transcriptional machinery, such as the TFIIA general transcriptional factor (Kaiser et al., 1995). We asked whether H-REV107-1 might potentially co-operate with STAT1 and BRCA1 in the IFNγ-response through a formation of a multiprotein complex including STAT1, BRCA1, PC4, and H-REV107-1.

We precipitated a multiprotein complex, consisting of H-REV107-1, PC4, and STAT1 proteins from COS-7 cells transiently transfected with the appropriate plasmids. However, the majority of the PC4 and STAT1 proteins remained in the protein extract, and only a minor fraction was bound to the H-REV107-1 (Fig. 17). We failed to express BRCA1 protein in COS-7 cells. Therefore potential binding of BRCA1 to the protein complex consisting of STAT1, PC4, H-REV107-1 proteins is unclear.

Summarising our investigation of IFNγ-signaling in the two human ovarian carcinoma cell lines OVCAR-3 and A27/80, we conclude that there are two different mechanisms of IFNγ - mediated growth suppression. In OVCAR-3 cells IFNγ leads to cell death trough the activation of STAT1, IRF1, H-REV107-1 signaling (Fig. 37). In A27/80 cells IFNγ leads to a cell cycle inhibition through STAT1, p21WAF1, and, hypothetically, BRCA1 activation. Possibly, in this pathway a STAT1, PC4, H-REV107-1 protein complex with or without BRCA1 protein is involved.

Hypothetical scheme of IFNγ-dependent growth suppression in OVCAR-3 and A27/80 cells is depicted in the Figure 37. There are several questions we wish to answer in our future investigations. Does H-REV107-1 interact with PC4 and other proteins the nucleus, and when does a transfer of the H-REV107-1 protein from the cytoplasm to the nucleus take place?. What is the functional role of the PC4, H-REV107-1, STAT1, and, supposedly, BRCA1 complex in IFNγ - signaling? Does H-REV107-1 influence STAT1 and PC4 activity as transcriptional activators in these cells, and is it involved in the regulation of expression of IFNγ - target genes.

[page 106↓]

Fig. 37 Hypothetical scheme of H-REV107-1 participation in the IFNγ-signaling in OVCAR-3 and A27/80 cells

4.3. H-REV107-1 Participates in the Cross-Talk between Retinoic Acid and IFNγ-Dependent Pathways

Retinoids (vitamin A and its metabolites) modulate cell growth, differentiation, proliferation, and can act as chemopreventive and chemotherapeutic agents for several types of cancer (Hong and Sporn, 1997). Notably, one of the H-REV107 – related proteins, lecithin retinol acyltransferase (LRAT), plays an important role in retinoid metabolism. It catalyses the esterification of retinol, thereby synthesising retinyl ester, which is supposed to act as a storage form for retinol in epithelial cells in skin, and breast (Chen et al., 1997). Investigation of the retinol metabolism in normal and cancer cells revealed that cancer cells have a greatly reduced ability to metabolise retinol into retinyl ester, This correlated with a significant reduction of LRAT expression (Guo and Gudas, 1998), suggesting that lack of retinyl esters in carcinoma cells may contribute to their tumorigenic phenotype (Guo et al., 2000).

It is known that the biologic activity of retinol and retinoic acid metabolism is mediated by retinoid binding proteins (CRBPs) and diverse retinoid nuclear receptors. There are two families of receptors: retinoic acid receptors (RARs) and retinoid X receptors (RXRs). Each family consists of three receptor types, alpha, beta, and gamma (Zhang et al., 2000).

[page 107↓]

They modulate gene expression by binding to specific DNA sequences (retinoid responsive elements, RARE) in the promoter regions of retinoid-target genes (Naar et al., 1991).

We identified the retinoic acid receptor gamma (RARG) in the yeast two hybrid system as a potential interacting partner of H-REV107-1. Interaction between RARG and H-REV107-1 proteins was a rather surprising finding, because this was a first evidence indicating a potential participation of the H-REV107-1 protein in retinoid signalling. Investigation of the interaction in COS-7 cells demonstrated that the H-REV107-1 protein does interact with RARG only in the presence of supplementary ligands.

To stimulate this interaction we used all-trans retinoic acid (ATRA), a synthetic ATRA antagonist (TTNPB), and a doublestranded sequence of the retinoic acid response element, DR5, which is known to be a target sequence of the RARs in promoter regions of the retinoid target genes (Idres et al., 2002). Although ATRA did not influence the interaction between RARG and H-REV107-1, TTNPB and DR5 weakly enhanced the protein binding. The specificity of binding was tested using a negative control with the mutated DR5 sequence. No enhancement of the interaction was detected (Fig. 32). Nevertheless, this interaction remained very weak, and only a small part of the intracellular H-REV107-1 was identified in the complex with RARG, a major part of the protein did not bind to the retinoid receptor (Fig. 32). Combining these data we suggest that under in vivo conditions the interaction might take place in the nucleus, where RARG bound to the promoter of target genes interacts with the H-REV107-1 protein. The weak interaction is likely to be due to the fact that only a small fraction of the H-REV107-1 protein exhibits a nuclear localisation.

It is still an open question if the H-REV107-1 – RARG interaction contributes to retinoid signaling. Recent investigations of RAR – binding proteins shed a new light at the known interaction partners of the H-REV107-1 protein. The STAT5 protein was identified as a critical regulator of enhanced RAR transcriptional activity (Si and Collins, 2002). Most interestingly, the authors described a direct overlap of STAT1 and STAT5-binding sites with RAR elements within the promoters of several genes, for example, RARβ and RARα (Langston et al., 1997). Such colocalisation of regulatory elements suggests a previously unexplored cross-talk between STAT and RAR families of transcriptional factors, following co-regulation of the IFN– and retinoid– dependent genes. Since the H-REV107-1 protein is a mediator of IFNγ-signaling (Sers et al., 2002), it will be definitely interesting to investigate the activity of the RARG in IFNγ-treated cells, and to define a potential role of IFNγ in the modulation of the RARG – H-REV107-1 interaction in more detail.

[page 108↓]

Another interesting aspect of the interaction between H-REV107-1 and RARG is the enhanced co-immunoprecipitation of the RARG and H-REV107-1 proteins in presence of the TTNPB ligand (Fig. 32). Binding to diverse ligands is supposed to change the conformation of the RARG molecule. Current models of the RAR activation suggest that binding of ligands results in a distinct conformational change, and leads to the release of co-repressors and the recruitment of transcriptional activators (Zhang et al., 2000), such as CBP, or p300. These factors are able to trigger RNA polymerase II to transcribe target genes (Kamei et al., 1996). Interestingly, another coactivator of the RNA polymerase II, the PC4 protein, was also found in the yeast two hybrid assay, as one of the H-REV107-1 binding proteins. Although a role of PC4 in the activation of RARs has not yet been described, its participation in this process cannot be excluded.

In summary, the functional relevance of the H-REV107-1, PC4 and STAT1 proteins on the RARG activity is still speculative, but provides an attractive hypothesis on the role of H-REV107-1 in both the IFNγ and retinoic acid signaling.

Importantly, IFN-γ was found to act synergistically with retinoids, and to enhance the growth inhibitory effect of retinoids in cultured breast and ovarian cancer cells (Hu et al., 2002). Analysis of the H-REV107-1 expression using cancer profiling arrays demonstrated down-regulation of the gene in breast, ovarian and lung tumors. IFNγ-treatment recovered H-REV107-1 expression in ovarian carcinoma cell lines OVCAR-3 and A27/80 (Sers et al., 2002). Investigation of the additive affect of IFNγ and ATRA revealed that IFNγ-treatment increased the expression level of RAR-alpha and RARG (Hara et al., 2001). Therefore it will be of interest to investigate if IFNγ-treatment of OVCAR-3 and A27/80 cells will lead to the up-regulation of RARG expression. Retinoids were demonstrated also to contribute to the cross talk between IFN- and ATRA – signaling pathways. In various cell lines, retinoic acid induces directly the expression of the transcription factors STAT1 and IRF-1, which play central roles in IFNγ-signal transduction (Chelbi-Alix and Pelicano, 1999). The H-REV107-1 gene was reported to be a target of IRF-1 (Sers et al, 2002). Possibly, H-REV107-1 participates in cross-talk between retinoid and IFNγ-signaling in ovarian tumors, although regulation of the gene occurs rather through a retinoid – independent pathway.

In contrast, the H-REV107-1 related protein, the H-REV107-2/TIG3/RIG1, is directly implicated in retinoid acid cellular response (Huang et al., 2000). The H-REV107-2/TIG3/RIG1 was cloned as an ATRA responsive gene in keratinocytes and cervical and gastric cancer cells, where it acts as a tumor suppressor (Di Sepio et al., 1998; Hung et al., 2000; Deucher et al., 2000).

[page 109↓]

Further investigation lead to the finding that in gastric carcinoma cells the H-REV107-2/TIG3/RIG1 protein induces apoptosis trough negative regulation of extracellular signal-regulated kinase, c-Jun N-terminal kinase (JNK), and p38 mitogen-activated kinase (Huang et al., 2002). Interestingly, other groups showed that treatment of these cells with ATRA leads to the inhibition of AP1 activity a subunit of which is the JNK-target protein, c-Jun. Investigation of retinoid receptors responsible for the ATRA-mediated growth inhibition of gastric cancer cells revealed that the retinoic acid receptor beta (RARbeta), but not RARG, is required for the AP-1 inhibition, and contributes to growth suppression (Huang et al., 2002). Taking these data together, we suppose that H-REV107-2/TIG3/RIG1 and RARbeta are involved in retinoic acid dependent growth arrest in gastric cancer cells, but not H-REV107-1 and RARG

4.4. H-REV107-1 – Mediated Cell Death through Inhibition of PP2A Activity

Protein phosphatase 2A (PP2A) is a major serine/threonine phosphatase in eukaryotic cells. PP2A can be considered a family of phosphatases, minimally containing a highly conserved catalytic subunit C (PR36) α and β isoforms; and a “scaffolding protein”, the regulatory subunit A (PR65), α and β isoforms. The isoforms of the PR65 subunit are 86% identical, and those of the PR36 subunit are 92% identical (Hemmings et al., 1990). The PR65 and PR36 proteins are tightly associated with each other, comprising the so called “core domain”. (Ruediger et al., 1994; Millward et al., 1998). The activity of the “core domain” is regulated by regulatory subunits B, bound to PR65. Four protein families B (PR55), B´ (PR61), B´´ (PR72, PR130, PR49, PR48), and B´´´ (PR93, PR110) are known as regulatory subunits B of PP2A (Table 8; Janssens and Gors, 2001).

Biological and genetic studies have revealed that PP2A enzymes are abundant, ubiquitous, and very conserved during evolution. PP2A dephosphorylates a large number of substrates in vitro, and is involved in the regulation of nearly all cellular activities, such as metabolism, transcription, cell cycle, signal transduction, and oncogenic transformation (Sontag, 2000). The PR65β encoding gene, PPP2R1B was characterised as a putative tumor suppressor down-regulated or mutated in lung, breast, and colon human tumors (Wang et al., 1998; Janssens and Goris, 2001). PR65α plays a critical role in cell morphogenesis of yeast, probably through regulation of the cytoskeletal network and cell wall synthesis. Deletion of the PR65α encoding gene, PPP2R1A is lethal in yeast (Kinoshita et al., 1996). Investigation of PR65α in human tumors demonstrated that PPP2R1A gene, is mutated in lung and breast human carcinomas and melanomas (Calin et al., 2000), strengthening a potential role of PP2A in human tumorigenesis.

The homozygous PR36α null mice are embryonically lethal, demonstrating that the PR36α subunit gene is essential for viability, and cannot be replaced by PR36β (Gotz et al., 1998). Recent data suggest that embryonic lethality results from defects in cell adhesion caused by insufficient levels of membrane-associated E-cadherin and β-catenin, suggesting a role of PP2A in Wnt/β-catenin – signaling (Seeling et al., 1999; Patturajan et al., 2002). The importance of PP2A in human cells is indicated by the fact that various pathogenic viruses, namely human immunodeficiency type 1 virus, HIV-1, (Tung et al., 1997), papova viruses (Pallas et al., 1990) or adenoviruses (Kleinberger and Shenk, 1993) express PP2A interacting proteins, which modify the phosphatase activity. Human cytomegalovirus (CMV), a herpesvirus, carries host-derived PP2A, associated with the nucleocapsid fraction (Michelson et al., 1996).

We demonstrated that the H-REV107-1 protein interacts with the regulatory subunit A α of the protein phosphatase 2A (PR65α) in COS-7 cells (Fig. 18), and under cell free conditions (Fig. 20). The interaction of the bacterially expressed H-REV107-1 with PR65α indicated that H-REV107-1 directly binds this protein. Co-immunoprecipitation assays in COS-7 cells with H-REV107-1 mutants delineated a region of 10 aminoacid residues on the very N-terminal end of the protein necessary for the interaction (Fig. 24).

Thus we have presented data demonstrating that the H-REV107-1 mediated apoptosis in OVCAR-3 cells is due to the inhibition of the PP2A activity through physical interaction between H-REV107-1 and the PR65 regulatory subunit. A challenge of our investigation was to define PP2A target proteins whose dephosphorylation plays such important role in cell survival.

[page 111↓]

Ectopically expressed H-REV107-1 protein forms homodimers in COS-7 cells. We tested the H-REV107-1 mutant proteins for a homodimer formation and observed that the same region responsible for the interaction with PR65α is also required for a homodimer formation. This suggests a competitive character of these interactions. Preliminary experiments demonstrated that binding to the PR65α protein excludes generation of homodimers (data not shown).

We wished to investigate a molecular mechanism of the correlation between H-REV107-1 mediated cell death and the interaction with PR65α. We observed that H-REV107-1 overexpression in human ovarian carcinoma cells OVCAR-3 results not only in the induction of apoptosis, but leads also to the transport of PR65α protein from nucleus into the cytoplasm (Fig. 27). During overexpression of the ΔC107-ΔN interaction deficient mutant neither apoptotic morphology of nuclei, nor transportation of PR65α were obtained. This suggested that H-REV107-1 – mediated apoptosis correlates with its ability to interact with the PR65α protein.

Little known about nuclear PR65α and its interacting partners in ovarian carcinoma cells. In most tissues and cell lines a cytoplasmic localisation of this protein was reported (Thou et al., 2003). We asked where the catalytic subunit PR36 is localised in OVCAR-3 cells, and if other PP2A subunits might interact with the nuclear PR65α protein. For example, the PP2A B56 regulatory subunit γ and δ isoforms were also shown to be concentrated in nucleus (McCright et al., 1996), suggesting that B56 isoforms might somehow be involved in this process. Preliminary analysis of expression of the genes encoding the B56 α, β, γ - proteins via RT-PCR revealed that all isoforms are expressed in OVCAR-3 cells (data not shown).

First, we investigated if H-REV107-1 can influence the intracellular localisation only of PR65α, or of the “core domain” consisting of PR65 and PR36 subunits. Immunofluorescence analysis of the intracellular distribution of PR36α in OVCAR-3 cells revealed an exclusively cytoplasmic localisation. Expression of the H-REV107-1 protein did not have any influence on the localisation of the PR36α protein (data not shown). PR36 is usually bound to the PR65 protein, and does not exist in a cell in free state. Therefore we supposed that the second isoform of the PR65 protein, PR65β, might be expressed in OVCAR-3 cells.

We performed a subcellular fractionation of OVCAR-3 cells, and tested the different fractions with an antibody recognising both, the PR65α and PR65β proteins. Western blot analysis revealed a strong specific signal in the cytoplasmic and in the nuclear fractions (data not shown). Therefore, we suppose that the PR65β is expressed in OVCAR-3 cells, and preferably localised in the cytoplasm, whereas the PR65α is localised in the nucleus, as we observed using immunofluorescence analysis.

[page 112↓]

Thus, it is likely that in the cytoplasm the PR36α - PR65β protein complex might exist. Taking into consideration that PR65α and PR65β are 86% homologue, we would like to test in our further experiment if H-REV107-1 might interact with PR65β in the cytoplasm, and if the PR36α, PR65β, H-REV107-1 complex exists in OVCAR-3 cells.

We determined a basal level of the protein phosphatase 2A enzymatic activity in untransfected OVCAR-3 cells, and in the cells over-expressing H-REV107-1. As a negative control we used a phosphatase inhibitor, okadaic acid, in the PP2A specific concentration of 2 nM (Holmes et al., 1990). H-REV107-1 overexpression had a strong inhibitory effect on PP2A activity, whereas the interaction deficient ΔC107-ΔN mutant was unable to inhibit the phosphatase (Fig. 30). This suggests that the interaction between H-REV107-1 and PR65α leads to the inactivation of the PP2A enzymatic activity.

The finding that PP2A inhibition leads to apoptosis is rather surprising, because it is in contrast with many other reports, suggesting that the protein phosphatase 2A is a pro-apoptotic phosphatase (Klumpp and Krieglstein, 2002). Additionally, OA has been described as a tumor promoting agent (Afshari et al., 1993; Fujiki and Suganuma, 1993).

A dual role of the protein phosphatase 2A in tumor progression has been already mentioned by other authors. Although most PP2A inhibitors are implicated in tumor promotion, the inhibitor fostriecin displayed a significant anti-tumor activity in Chinese hamster ovary cells inducing a dose-dependent arrest of cell growth during the G2-M phase of the cell cycle (Cheng et al., 1998). Recent research in Drosophila melanogaster demonstrated that a trimeric complex PP2A/C, PR65, and B56-regulatory subunit is required for survival and protect cells from apoptosis (Li et al., 2002). Investigation of the sensitivity of different cell lines to OA-induced apoptosis revealed an unexpected result. It was shown that the cell lines harbouring ras mutations are more sensitive than the cell lines with wild-type ras (Rajesh et al., 1999). Using dual-coloured flow cytometry, we demonstrated that treatment of OVCAR-3 cells with 10 nM OA results in cell death due to apoptosis (Fig. 30). Although OVCAR-3 cells do not carry Ras mutations (Patton et al., 1998), activation of one of the Ras-dependent down-stream pathways, the PI3-kinase pathway, has been reported in these cells (Shayesteh et al., 1999). PI3-kinase plays an important role in a cell survival. Amplification of its downstream kinases Akt1, and Akt2 was demonstrated in human ovarian cancer, and inhibition of this pathway led to the cell death (Yuan et al., 2000).

Therefore we included the PI3-kinase inhibitor, LY294002 in further analysis, to test if LY294002 and OA treatment have a similar effect on the OVCAR-3 cells, like H-REV107-1 ectopic expression. Investigation of the apoptotic pathway after H-REV107-1 overexpression, LY294002, and OA treatments revealed that a mitochondrion-dependent, but not a death receptor – dependent pathway is activated in all three cases.

[page 113↓]

Cleavage of the caspase-9, and its down-stream caspases-3 and –7 was detected (Fig. 31; and data not shown). In contrast, a marker for the death-receptor dependent apoptosis, cleaved caspase-8 was revealed only after treatment with 10 nM OA (data not shown), suggesting that the cellular response to higher concentrations of OA includes activation of additional targets, such as PP1, which activates a death receptor pathway, that differs from the response to the H-REV107-1 protein. Apoptosis induction after treatment with 10 nM OA, and after H-REV107-1 overexpression was more pronounced as after treatment with 0.5 nM OA (Fig. 31, B), probably due to the insufficient inhibition of PP2A activity by low concentration of okadaic acid. Although PP2A is inhibited by OA with Ki 0.2 nM (Shima et al., 1994), various cell lines differ in their membrane’s penetrance. Therefore it is difficult to control intracellular OA concentration, and some unspecific effects can occur.

We have analysed PP2A substrate proteins, such as members of the BCL-2 family, directly participating in mitochondrion-dependent apoptosis, and proteins involved in RAS.-dependent signaling. One of the proapoptotic members of this protein family, is the BAD protein function of which is regulated by reversible phosphorylation. Three phosphatases, PP1 (Ayllon et al., 2000), PP2A (Chiang et al., 2001), and PP2B (Wang et al., 2000) have been demonstrated to be implicated in the dephosphorylation of the Ser/Thr residues within BAD. Inhibition of phosphatase activity by treatment of cells with okadaic acid of different concentrations demonstrated that in OVCAR-3 cells BAD is rather a target of PP1 phosphatase but not of PP2A (data not shown). Analysis of other members of the BCL-2 family, Bcl-2 and Bxl revealed negative results as well.

The RAS-downstream signaling pathways were of our particular interest. Two pathway down-stream of Ras, the RAF-MEK-ERK and the PI3K-Akt pathway have been described to be regulated via PP2A (Millward et al., 1999). PP2A modulates the activity of several members of the RAF-MEK-ERK pathway, and has both positive and negative effects (Frost et al., 1994; Alessi et al., 1995; Zhou et al., 2002).

PP2A dephosphorylates an inhibitory site within Raf, Ser-259, the dephosphorylation of which was shown to be necessary for full activation of this kinase (Abraham et al., 2000). We analysed activation-phosphorylation of ERK, MEK, and RAF in OVCAR-3 cells treated with 1 nM and 10 nM OA, with LY294002, and transfected with an H-REV107-1 expression vector. We did not obtained any effect on the phosphorylation of RAF1. Treatment with LY294002 and 1 nM OA led to the reduction of the phosphorylated RAF-down-stream kinase MEK (Nazarenko, unpublished data), suggesting a cross-talk between PI3-kinase and RAF-MEK-ERK kinase pathways. One of the PI3-kinase down-stream effectors, the PKC kinase, has been reported as an activator of MEK (Kolch, 2000).

[page 114↓]

A large number of studies demonstrated a PP2A inhibitory effect on the PI3-kinase – dependent pathways (Sontag, 2000; Yellaturu et al., 2002). Akt, and PKC were described as substrates of PP2A (Yellaturu et al., 2002; Sontag et al., 1997; Standaert et al., 1999). Interestingly, that PKC revealed an inhibitory effect on Akt - activity (Wen et al., 2002), indicating a negative feedback loop within a PI3-kinase pathway.

Thus, PP2A might activate the PI3-kinase pathway, through inhibition of the PKC activity, and inhibit the PI3-kinase pathway through inhibition of Akt activity. It is possible that different regulatory subunits of PP2A might recruit the enzyme to either PKC, or Akt kinase, and modulate therewith a cell survival or cell death, respectively.

Fig. 38 Schematic presentation of the mechanism of H-REV107-1 – mediated cell death

Inhibition of a pathway is shown with red lines, activation of a pathway is shown with green lines.

[page 115↓]

It is a desire of our further experiments to elucidate if the PKC kinase is a primary target of PP2A in OVCAR-3 cells. This might answer the question which activity of PP2A, essential for cell survival, inhibits H-REV107-1. A recent publication, demonstrating that PKCε is necessary for the IFNγ-response, supports our hypothesis that PKC might be a target of PP2A in OVCAR-3 cells (Ivaska et al., 2003). The authors demonstrate that the PKCε kinase phosphorylates, and thereby activates the STAT1 protein. PP2A is responsible for a deactivation of PKCε via dephosphorylation. This results in attenuation of the IFNγ-induced phosphorylation of STAT1, and, therefore, inhibition of IFNγ-signaling. It was demonstrated that that IFNγ - function might be recovered by inhibition of PP2A.

We did not yet prove if OA treatment will recover an IFNγ-response in OVCAR-3 cells, but the fact that only the cells over-expressing the H-REV107-1 protein, inhibiting PP2A activity, underwent apoptosis after IFNγ-treatment, substantiates this possibility. A hypothetical scheme of the PP2A signaling pathway in OVCAR-3 cells is depicted in the Figure 38.

4.5. Possible Participation of H-REV107-1 in Calcium Metabolism

One of the H-REV107-1 interacting partners that was confirmed in yeast and mammalian cells is the Ca2+ - binding protein calcyclin. Calcyclin belongs to the S100 family of calcium binding proteins, and was characterised as a marker for the malignant phenotype in human malignant melanomas (Maelandsmo et al., 1997 ), squamous cell carcinomas (Berta et al., 1997), breast (Pedrocchi et al., 1994), and colon cancers (Stulik et al., 2000). There is ample evidence that calcyclin and other members of this family play an important role in neoplastic progression (Weterman et al., 1992). Recent research supports these earlier findings, demonstrating that up-regulation of calcyclin in colorectal adenocarcinomas is directly correlated with Dukes’s tumor stage. This suggests that calcyclin may be involved in the progression and invasive process of human colorectal adenocarcinomas (Komatsu et al., 2000; Bronckart et al., 2001).

The interaction between H-REV107-1 and calcyclin is dependent on the presence of Ca2+ and Zn2+ ions in the lysis buffer (Fig. 33). These ions were demonstrated to bind different sites of calcyclin, independent of each other, changing the protein conformation, and probably the binding capacity of calcyclin (Kordowska et al., 1998). Therefore, it will be of a particular interest to investigate separately a role of these ions on the binding of calcyclin to the H-REV107-1 protein.

Until now we have no further evidences for a functional role of the H-REV107-1 protein in Ca2+ signaling. Interestingly, one of the regulatory subunits of PP2A, another H-REV107-1 interacting protein, has been described as a calcium binding protein, and a role of PP2A in calcium metabolism was suggested (Janssens et al., 2003).

[page 116↓]

The authors evaluated the effects of Ca2+ on subunit composition, subcellular targeting, and catalytic activity of a PR72-containing PP2A trimer. Binding of calcium required nuclear localisation of PR72, and mediated PP2A activity in vitro. H-REV107-1 was demonstrated as a negative regulator of PP2A activity in vitro, and its forced expression lead to the transport of the PR65 subunit form nucleus into the cytoplasm.

These data support the notion that the H-REV107-1 protein is directly involved in the calcium metabolism. Therefore it will be of a particular interest to investigate whether calcyclin is able to mediate the PP2A activity in OVCAR-3 cells in presence and absence of the H-REV107-1 protein, and to investigate a potential influence of the Ca2+ or Zn2+ ions on the H-REV107-1 – mediated inhibition of PP2A activity.

In summary, the application of the yeast two hybrid approach revealed multiple interacting partners of the class II tumor suppressor H-REV107-1, suggesting its participation in various cellular processes. We demonstrated that the H-REV107-1 protein is an important mediator of IFNγ signaling, and induces either growth suppression, or apoptosis through different mechanisms. The STAT1, PC4, H-REV107-1 protein complex is likely to be involved in IFNγ - mediated growth suppression in ovarian carcinoma cells.

Further studies identified a novel oncogenic role of the protein phosphatase 2 A (PP2A) in ovarian carcinoma cells. PP2A is a central serine/threonine phosphatase which is involved in nearly all cellular processes. We demonstrated that for the H-REV107-1 - mediated cell death, the interaction with the regulatory subunit A (PR65) of PP2A is required. The H-REV107-1 - PR65 protein binding results in the inhibition of PP2A activity, which is followed by the induction of apoptotic signaling via a caspase-9 – dependent pathway. The challenge of our future investigation will be to define the direct target of PP2A which is responsible for tumour cell survival, and to elucidate the mechanisms of H-REV107-1 mediated growth inhibition or cell death in more detail.

© Die inhaltliche Zusammenstellung und Aufmachung dieser Publikation sowie die elektronische Verarbeitung sind urheberrechtlich geschützt. Jede Verwertung, die nicht ausdrücklich vom Urheberrechtsgesetz zugelassen ist, bedarf der vorherigen Zustimmung. Das gilt insbesondere für die Vervielfältigung, die Bearbeitung und Einspeicherung und Verarbeitung in elektronische Systeme.
DiML DTD Version 3.0Zertifizierter Dokumentenserver
der Humboldt-Universität zu Berlin
HTML generated: