REVIEW OF THE LITERATURE

Unlimited growth, the most characteristic feature of tumors, represents the sum of unfavourable events. First, a genetic change, due to intrinsic or extrinsic signals, has taken place and secondly, recognition of DNA damage or its repair process has failed. Once genomic instability has arisen, additional changes accumulate more easily. Later, selection favours malignant cells with increasingly aggressive growth properties and allows rapid expansion of the tumor. Although a cell has diverse cautionary measurements to prevent loss of growth control, some of them may fail. Most commonly, cancer results from tumorigenic abberrations of cell cycle regulators, in particular, those governing the G1 progression and G1/S-transition.

Cells, tumor or normal cells, follow the steps of cell cycle consisting of G1, the actual growth phase during which they prepare to synthesize DNA; S, DNA replication; G2, a second shorter growth phase; and M, mitosis. Cells that have transiently or permanently exited the cell cycle, for example because of serum starvation, terminal differentiation, or senescence, remain in the G0 phase. Cells are constantly a target for mitogenic and antimitogenic signals that affect them only during the G1 phase. Growth factors are required for progression through G1 to a specific point called the restriction point (R-point), after which the cell is committed to complete the division cycle even in the absence of growth factors thereafter (Pardee, 1989) (Fig. 1).

The role of cyclins and cyclin-dependent kinases (Cdks) in the precise control of cell cycle is well established. Mitogenic signals cause the assembly of different kinase holoenzymes, that are composed of a cyclin regulatory subunit and a Cdk catalytic subunit. These complexes are formed and activated at specific phases of the cell cycle, and their coordinated function is required for progression through the S and M phases. Contrary to cell cycle phase-specific expression of cyclins, Cdks are expressed constitutively, but are not active until complexing with suitable Cdk. Additionally, the kinase activity of certain Cdks can be regulated by phosphorylation of critical residues, or kinase activity inhibited by various Cdk-inhibitors (CKI), some of which have been shown to possess tumor suppressive properties. p21, p27 and p57 are universal Cdk-inhibitors with broad specificity for Cdks, whereas the INK4 (inhibitor of Cdk4/6) family members p15, p16, p18, and p19 selectively inhibit only the kinase activity of Cdk4 and Cdk6 in the G1 phase (Kamb, 1995).

The tumor suppressor gene p53 and the retinoblastoma susceptibility gene (RB) are both involved in the regulation of the cell cycle and are closely linked to cyclin-Cdk pathway. While retinoblastoma protein (pRB) is required in the control of normal progression of the cycle, p53 is called into action in the case of genetic insult. pRB is phosphorylated in a cell cycle dependent manner and is in the hypophosphorylated, active form in the early G1 and binds transcription factor E2F-1. Phosphorylation of pRB by cyclin D-Cdk4/6 in the late G1 results in the release of E2F-1, which activates genes required for the progression to the S phase (Fig. 1.). At present, pRB is the only known target of cyclin-Cdk activity in G1.

The antitumorigenic potential of p53 correlates with its ability to check the integrity of the genome and cease the progression of the cell cycle when DNA damage is encountered. Upon genetic insult, p53 transactivates, among other target genes, the expression of p21 Cdk-inhibitor and thus participates in the growth arrest of cells (Fig. 3). p53, often called the guardian of the genome, is the most often mutated gene in human cancers, stressing the importance of p53 and cell cycle control in tumorigenesis. Whatever the mechanism, increased activity of cyclins and Cdks as well as decreased function of CKIs, pRB, and p53, highly predispose cells to transformation.

Cells are constantly exposed to both extrinsic and intrinsic DNA damage signals. Extrinsic sources of damage include irradiation and chemical mutagens, while intrinsic damage is generated by the cell itself. Despite the high spontaneous chemical reactivity of DNA, only one error is made during replication of the 3x10⁹ bases of human genome. Proliferating cells have a great capacity to tolerate genomic errors, as cells can repair about 100 spontaneous alterations per hour (reviewed by Lehmann and Carr, 1994).

DNA damage types can be grouped into two main categories: damage modifying nitrogenous bases, and damage alterating the phosphodiester backbone. Chemical agents can be covalently attached to bases making intrastrand or interstrand cross-links or cause depurination or depyrimidination of DNA (reviewed by Friedberg, 1995). Ultraviolet (UV) radiation causes special kinds of changes: cyclobutane pyrimidine dimers and (6-4) photoproducts. These lesions are repaired by photoreactivation of DNA and by nucleotide excision repair (NER). Most DNA lesions produced by free oxygen radicals are corrected by base excision repair which is distinct from NER. In this repair system the damaged base is cleaved from its deoxyribose moiety and the apurinic or apyrimidic (AP) sites are excised followed by DNA repair synthesis and ligation. AP sites can also result from spontaneous depurination or depyrimidination of DNA. Mismatch repair, on the other hand, refers to correction of mispaired bases, which are frequently produced during replication and recombination. It can occur by several biochemical pathways including base excision (reviewed by Friedberg, 1995). Broken phosphodiester bonds, on the other hand, cause double-stranded breaks which are repaired by nonhomologous end-joining or by the addition of new telomeres (Wilkie et al., 1990).

Cells have a number of systems to interrupt cell cycle progression when damage to the genome is detected or when cells have failed to complete the preceding cell cycle phase. The DNA damage checkpoint can be regarded as a signaling system, where information flows from factors detecting DNA lesion to cell cycle targets. Signals activating the checkpoints can be produced from recognition of DNA damage itself, during repair, or during replication of damaged sites (Lehmann and Carr, 1994). DNA damage checkpoints act at the G1/S border, during S, and at the G2/M boundary. Although arrest points are scattered throughout the cell cycle, the same or similar proteins are often involved in all of them. In addition, many of these proteins function not only in checkpoint control but also in DNA repair, apoptosis, and transcriptional induction (Paulovich et al., 1997). p53, for example, has a role in each DNA damage checkpoint, in spindle checkpoint, and in apoptosis.

The desired outcome of DNA damage checkpoint control is either a successful repair of DNA after growth arrest, or, in the case of permanent damage, induction of apoptosis. In both these situations DNA errors are not mediated onwards, whereas replication of damaged DNA leads to accumulation of mutations or genomic instability. In addition to the intrinsic error rate, the checkpoint control system may fail because of adaptation or selective pressure. In adaptation, the cell cycle continues after an interval of arrest even if the damage remains unrepaired (Sandell et. al., 1993). In some pathological conditions, like cancer, selection favours cells with defective checkpoints, since genomic instability gives rise to multiple genetic changes needed both for the primary transformation and for increased invasiveness (reviewed by Paulovich, 1997).

UV radiation can be divided into three main classes according to its wavelength; UVA (320-400 nm), UVB (290-320 nm), and UVC (100-290 nm). Because the atmospheric ozone layer efficiently absorbs the shortest waves of UV radiation, solar light consists only of UVA and UVB. Even if human skin never meets UVC light, the same DNA lesions, although at a lower efficiency, are produced after longer wavelengths of UV. Besides typical UV-induced lesions, UVA/B also causes considerable oxidative damage. Being usually easily available and measurable, UV radiation, especially UVC, has been quite extensively used in studies of DNA damage and repair. When DNA damage studies are evaluated, the nature of genetic insult is of great importance, as the damage response differs between DNA damage types.

UV radiation causes the formation of a specific type of DNA lesions, the "UV footprints". When DNA is exposed to UV, adjacent pyrimidines become covalently linked by the formation of four-membered ring structures, referred to as cyclobutane pyrimidines or pyrimidine dimers. These lesions distort the helical structure of DNA, and some isomeric forms, but not all, are able to obstruct DNA replication and transcription (Taylor et al., 1990). Formation of pyrimidine dimers is a reversible process, and with high UV doses an equilibrim between the formation and dissociation of dimers by photoreversal can be reached. The dimer formation seems not to take place randomly between pyrimidines, but thymine containing dimers are preferred over cytosine containing ones. With UVC of 254 nm the ratio of T-T to C-T to T-C to C-C was found to equal 68:13:16:3 (Tornaletti et al., 1993; Mitchell et al., 1992).

Another typical UV-induced DNA lesion is the formation of pyrimidine-pyrimidone (6-4) photoproducts, most often of the TC and CC types, which considerably distort the DNA helix (Taylor et al., 1988). Generally the incidence of photoproducts after UV radiation is several times lower than that of pyrimidine dimers, although at occasional sites in DNA (6-4) lesions may occur as frequently as dimers (Bourre et al., 1987) . However, the frequency of UV-induced nonsense mutations correlates better with the frequency of photoproducts than of pyrimidine dimers, suggesting that (6-4) lesions may have a stronger contribution to the mutagenicity of UV radiation (Mitchell et al., 1989).

Although UV and g-radiation seem to cause quite similar cellular responses, growth arrest and apoptosis, the nature of DNA damage is quite different. The most striking lesions made by ionizing radiation are direct single and double strand breaks, which are almost never caused directly by UVC. However, DNA-protein cross-links and DNA strand breaks are sometimes detected at longer wavelengths of UV (Tyrrell, 1991).

Nucleotide excision repair is the most extensively studied, universal, and highly conserved DNA repair mechanism, which is used in all cell types. Many types of DNA damage, particularly those produced after UV radiation or exposure to chemical agents, are repaired by NER. As in base excision repair the damaged base or nucleotide is excised and replaced with a newly synthesized DNA strand by using a complementary strand as a template. An enzyme system hydrolyzes two phosphodiester bonds, one at either side of the damage, to generate an oligonucleotide carrying the lesion. After removing the oligonucleotide, the gap is filled by repair synthesis and ligated to the existing strand (reviewed by Frieberg, 1995). Upon UV radiation, DNA strand breaks are formed only in the context of repair, not directly by UV. Rationally enough, NER occurs more efficiently in the leading strand of DNA and in actively transcribed genes (Mellon et al., 1987) . Moreover, some components of TFIIH, a factor essential for transcription initiation, are also required for NER suggesting that repair and transcription are not fully separate processes but are linked to each other and share, at least partly, the same components (Bootsma and Hoeijmakres, 1993). For example, in NER several DNA unwinding enzymes called helicases are needed for damage recognition and removal of a damaged segment, and some of these helicases, ERCC2 and ERCC3, are also components of TFIIH complex thus acting both in nucleotide excision repair and in transcription initiation (Bootsma and Hoeijmakres, 1993).

Three distinct hereditary diseases are associated with defects in nucleotide excision repair: xeroderma pigmentosum (XP), Cockayne´s syndrome (CS), and trichothiodystrophy (TTD). Although in all syndromes gene(s) involved in NER is/are mutated, the clinical pictures are quite diverse. Contrary to expectations, only XP patients are prone to cancers. They exhibit many sun light-induced diseases, including skin cancer, and some neurological defects (Bootsma, 1993). In fact, discovery of defective NER in XP patients led to subsequent cloning of nuclear excision repair genes, which were named XP genes or ERCCs (excision repair cross complementing). XP patients have been assigned to seven different groups (XP-A through XP-G), each carrying a mutation in a different gene. In Cockayne´s syndrome, the CSB (=ERCC6) and XP-D (=ERCC2) genes are mutated, resulting in a unique repair defect; a lack of the coupling of transcription for repair (van Hoffen et al., 1993). TTD is caused by mutations of at least XP-B (=ERCC3), XP-D (=ERCC2), and XP-G (=ERCC5) genes (Stefanini et al., 1993). Although XP patients have a high incidence of UV-induced malignacies, there is no other evidence of malfunction of NER in cancers, besides this rare syndrome.

Children with familial retinoblastoma carry a germline mutation on one allele of the retinoblastoma gene (RB1), and mutation of the other allele later gives rise to clinical cancer at very young age (Cavenee et al., 1988). It was soon discovered that RB was mutated in several other cancer types as well, and the identification and characterization of the RB gene made it the first tumor suppressor cloned.

Rather than a tumor suppressor only, pRB should be considered as a core element of an essential pathway governing cell cycle progression, differentiation, cell death, and tumorigenesis. This "RB pathway" consists of D-type cyclins and Cdk4 and 6, which are largely responsible for the cell cycle phase-specific phosphorylation and inactivation of pRB, p16 CKI inhibiting the activity of those kinase complexes, pRB itself, and the family of E2F transcription factors regulated by pRB (Fig 2.). The intactness of this pathway seems to be of utmost importance in tumor suppression, since one or more of these cell cycle regulators appear to be aberrant in nearly every tumor (reviewed by Sherr, 1996). Activation of components with oncogenic potential (cyclin-Ds, Cdk4/6, cyclin E, E2Fs), or inactivation of tumor suppressors (CKIs, pRB) may lead to uncontrolled cell proliferation. Depending on cancer type, a variety of different molecular mechanisms are detected in deregulation of different components of this pathway.

Phosphorylation of pRB

The main task of pRB is to control the progression of the cell cycle through late G1 and the commitment to enter S. pRB is periodically phosphorylated on its serine and threonine residues; at least a dozen phosphorylation sites have been identified (Knudsen and Wang, 1996). Unphosphorylated pRB has to exert its functions during the first two thirds of the G1 phase, during which the cell is sensitive to mitogenic signals. Around and after this restriction point pRB becomes phosphorylated, and is in this form unable to influence the course of the cycle until it is dephosphorylated again after M phase (Ludlow et al., 1990)(Fig.1.). Cyclin D1, 2, and 3 together with Cdk4 and Cdk6 are most prominently involved in the phosphorylation of pRB, but overexpression of cyclin E and A also enhances the phosphorylation of pRB in vitro (Hinds et al., 1992). Probably cyclin Ds are responsible for pRB phosphorylation during G1-phase progression, while kinases associated with cyclin E/A take over the phosphorylation of newly synthesized pRB at the G1/S transition and thereafter (Weinberg, 1995; Bartek et al., 1996). Cyclin D1 overexpression often seen in squamous cell carcinomas is mainly due to gene amplification, whereas in mantle cell B lymphomas an increased level of cyclin D1 is caused by cromosomal translocation t(11;14) (Bartek et al., 1997).

The activity of pRB is modulated in response to both positive and negative growth signals: mitogens favoring cell proliferation raise the level of cyclins leading to phosphorylation of pRB, whereas growth inhibitory signals, such as transforming growth factor b (TGF-b and contact inhibition, modulate the activity of Cdks by CKIs resulting in inhibition of pRB phosphorylation (Reynisdottir et al., 1995). Thus, neither signal types affect pRB directly, but through modulating the function of cyclin-Cdk complexes.

Complex of pRB with E2Fs

pRB exerts its effects by interacting cell cycle phase-dependently with a range of factors regulating proliferation and differentiation. The growth-inhibitory function of hypophosphorylated pRB was thought to be based on its ability to bind and sequester E2F transcription factors (Chellappan et al., 1991)(Fig. 2.). Later, the pRB-E2F complex was found to actively reppress gene transcription (Hamel et al., 1992; Weintraub et al., 1992), and a recent piece of evidence indicates that the active repression by E2F-pRB complex, not the inactivation of E2F, is the prerequisite for pRB-mediated G1 arrest (Zhang et al., 1999). pRB is a member of the "pocket-protein" family, as a large A/B "pocket" consisting of A and B protein-binding domains is responsible for binding E2Fs and many other factors. The E2F family comprises six transcription factors (E2F1-5 and dE2F), which heterodimerize with members of DP family and control the expression of a number of S-phase genes containing E2F-responsive elements in their promoters (reviewed by Dyson, 1998). The growth inhibitory function of pRB is not limited to E2F repression-mediated inhibition of RNA polymerase (pol) II, but pRB can also inhibit the activity of pol I (Cavanaugh et al., 1995) and pol III (White et al., 1996). Moreover, pRB is able to repress the transcription of Myc (Cziepluch et al., 1993) and Fos via retinoblastoma controlling element and to enhance the transcription of insulin-like growth factor II (IGFII).

Excess RB has been reported to protect cells from apoptosis, probably by inhibiting E2F-mediated apoptosis (Haas-Kogan et al., 1995; Haupt et al., 1995; Berry et al., 1996). E2F-1 overexpression drives cells inappropriately into the S phase (Johnson et al., 1993) and causes apoptosis, supposedly in a p53 dependent manner (Qin et al., 1994; Shan and Lee, 1994; Wu and Levine, 1994). Interestingly, expression of p19ARF, an activator of p53, is induced by E2F-1 providing a link between the pRB and p53 pathways (DeGregori et al., 1997). E2F-1 seems to be a protein of dual nature. On one hand, it behaves like a classic oncogene in transformation assays (Johnson et al., 1994) and its overexpression promotes tumorigenesis in transgenic mice (Pierce et al., 1998). On the other hand, E2F-1 knock-out mice suffer from numerous tumors (Yamasaki et al., 1996). These opposite effects seem to be dependent on the concentration of E2F in a specific manner: high levels promote apoptosis, medium levels allow cell cycle progression, and low levels cause growth arrest. It is particularly the capacity of E2F-1 to induce apoptosis that seems to be relevant for its conditional tumor suppression function (Macleod, 1999).

The inhibition of E2F has turned out to be too simplistic an explanation for the tumor suppressing properties of pRB. In contrast, it has been shown that E2F binding is not sufficient for growth suppression by pRB (Qian et al., 1992). Moreover, amplification of E2F in cancer cells is extremely rare and has been found only once (Saito et al., 1995). However, tumors in RB +/- mice are smaller and are formed later in the absence of E2F (Yamasaki et al., 1998). To date, pRB has been found to bind to at least 50 different proteins, and clearly some of these interactions may have a significant role in RB-mediated tumor suppression. For example, the C-terminal domain of pRB binds mdm2 (Xiao et al., 1995) and c-Abl tyrosine kinase (Welch and Wang, 1995), both involved in regulation of tumorigenesis. By binding multiple effectors pRB can simultaneously modulate the signals from diverse growth controlling pathways.

The balance between RB and p16

The inverse correlation between pRB and Cdk-inhibitor p16 expression that is observed in many cancer types indicates that they are connected to each other by a negative feedback loop. As a G1 phase specific Cdk-inhibitor, p16 is mainly involved in the suppression of pRB phosphorylation by cyclin D-Cdk4/6 complexes.

One of the first observations indicated that p16 was overexpressed in tumor cells harboring oncoprotein-inactivated pRB (Serrano et al., 1993). Subsequently, the loss of RB function was associated with high levels of p16 in many tumor types, particularly in gliomas and bladder carcinomas. The introduction of wild-type p16 arrests normal cells in late G1, a mechanism requiring functional pRB (Lukas et al., 1995; Serrano et al., 1995). The ability of pRB to negatively regulate p16 promoter explains the high levels of p16 in the absence of functional RB (Li et al., 1994). Conversely, p16 has been shown to induce transcriptional downregulation of RB, completing the feed-back loop (Fang et al., 1998).

Point mutations and small deletions of p16 are common in pancreatic adenocarcinomas, esophageal carcinomas, and biliary tract cancers, and germline mutations of p16 are found in hereditary melanomas and pancreatic cancers (reviewed by Pollock et al., 1996). Other possible inactivating mechanisms include homozygous deletions and methylation of the p16 locus. p16 Cdk-inhibitor is also the most often lost cell cycle regulator in established cell lines; 70% of immortalized cell lines contain no functional p16.

In addition to disturbing RB function by abnormal regulation of its phosphorylation, pRB activity can be prevented by point mutation or deletion of RB gene, or by binding of pRB with viral oncoproteins. The protein products of transforming DNA viruses, such as simian virus 40 (SV40) T antigen (DeCaprio et al., 1988), human papillomavirus (HPV) E7 protein (Dyson et al., 1989), and adenovirus E1A protein (Whyte et al., 1988) are able to bind to the hypophosphorylated form of pRB and prevent its association with E2F. Elimination of pRB activity is necessary but not sufficient for virus induced transformation, and explains why the same viruses, though through different proteins, also target p53.

Besides being defective in all retinoblastomas, inactive RB alleles are encountered in approximately 90% of small cell lung carcinomas (Harbour et al., 1988; Yokota et al., 1988; Horowitz et al., 1990), in 20-30 % of non-small cell lung cancers (Reismann et al., 1993), and frequently in bladder, pancreatic (Ruggeri et al., 1992), prostate, breast (Lee et al., 1988; T´Ang et al., 1988; Bookstein et al., 1989), and mesenchymal cancers (Friend et al., 1986; Weichselbaum et al., 1988). Patients with familial retinoblastoma are also prone to soft tissue sarcomas and osteosarcomas (DerKinderen et al., 1988). While major deletions are responsible for defective RB in retinoblastomas, inactivating point mutations in the pRB pocket are more common in other cancers. Interestingly, high levels of pRB in bladder cancer has been associated with as poor a prognosis as is loss of pRB, and overexpression of pRB was found to correlate with loss of p16 further confirming the importance of the intactness of the whole RB pathway (Benedict et al., 1999).

RB appears to be indispensable for development, because RB-/- mouse embryos die at midgestation due to defective erythropoiesis and uncoordinated proliferation and cell death in the liver, lens and nervous system (Clarke et al., 1992; Jacks et al., 1992; Lee et al., 1992). Although RB is clearly important in cellular differentiation, RB knockout mice develop normally until the 13th gestational day, and RB-/- mouse embryo fibroblasts (MEFs) derived from RB-/- mice before this stage are able to proliferate but have a somewhat shorter G1 phase than normal MEFs (Herrera et al., 1996). Other RB family members, p107 and p130, may partly take over the functions of pRB and explain the normal growth at very early stages of development. RB heterozygous mice are viable and develop pituitary and thyroid tumors, but contrary to expectations, exhibit no retinoblastomas (Hu et al., 1994; Nikitin and Lee, 1996). However, RB+/- p107 -/- mice show retinal dysplasia (Lee et al., 1996) or retinoblastoma (Robanus-Maandag et al., 1998) depending on the mouse strain, suggesting that due to overlapping functions inactivation of both pRB and p107 is needed for retinal phenotype. Simultaneous inactivation of pRB and p53 also predisposes to retinal tumors; p53 function would otherwise commit the cells to apoptosis (Howes et al., 1994).

Originally, p53 protein was found to complex with SV40 large T antigen (Lane and Crawford, 1979; Linzer and Levine, 1979), and its overexpression appeared to cause oncogenic transformation of cells. These findings misled the first investigators to regard p53 as a tumor antigen or an oncogene. Later, when the wild-type conformation of p53 was found, the true nature of p53 was revealed. Wild-type p53 was often found inactivated in human tumor cells (Wolf and Rotter, 1985; Baker et al., 1989; Kelman et al., 1989; Nigro et al., 1989) and, when introduced into cells, was growth suppressive rather than oncogenic (Finlay et al., 1989; Baker et al., 1990; Michalowitz et al., 1990).

p53 mutations

p53 function is lost in over 50% of human cancers evidencing the role of p53 in tumorigenesis (Hollstein et al., 1991). A recent database contains over 7500 p53 mutations in all human tumors (Beroud and Soussi, 1998). The mutation frequency varies from one cancer type to another; the average mutation frequencies in most common malignancies are 70% in lung carcinoma, 65% in colon cancer, 45% in stomach cancer, and 30% in breast and prostate cancers (Beroud and Soussi, 1998). Tumor types that rarely contain p53 mutations may still harbor some other changes inactivating p53 pathway. Nevertheless, some cancer types, such as teratocarsinomas, seem to be highly resistant for selecting p53 mutations (Lutzker and Levine, 1996). Most commonly the genetic change comprises a missense mutation in one allele, producing a faulty protein with an increased half life. The vast majority of these p53 mutations are clustered in the DNA-binding domain of the protein, particularly within the four evolutionary conserved regions, the so-called hot spots (Hollstein et al., 1994) (Fig.3). Point mutations can be divided into two classes according to the way they change the DNA-binding ability: "contact" mutants decrease p53 DNA-binding by interfering with those amino acid residues making contact between p53 protein and DNA, whereas "conformational" mutants, making up only 7% of p53 mutations, disrupt the structural elements of the protein. Instead of a specific, common mutant conformation of p53, mutant p53 is more likely a defectively folded form of the protein (Cho et al., 1994).

The mutants of p53 may achieve tumorigenic properties also with a mechanism other than losing the wild-type functions. Some p53 mutants can oligomerize with wild-type p53 and hinder the DNA-binding and transactivation functions of p53 through their dominant-negative effect. For example, p53 wild-type mice with a dominant-negative transgene developed tumors in which selection has favored mutant p53 over wild-type p53. Moreover, mutant p53 transgene was able to cause tumors in mice harboring one or two alleles of wild-type p53, but not in mice lacking p53, demonstrating the difference between dominant-negative effect and gain-of-function properties (Harvey et al., 1995). Although generally inactive, some mutant forms are still capable of binding DNA and transactivating p53 target genes, at least to a certain degree or in certain circumstances.

While some p53 mutants may mimic the actions of wild-type p53, some wild-type p53 proteins can be functionally inactive. The cytoplasmic location of wild-type p53 sometimes seen in certain rare forms of breast cancer, neuroblastoma (Moll et al., 1995), and melanoma (Weiss et al., 1995) renders p53 incapable of transactivating genes in the nucleus. High levels of mdm2, due to gene amplification or other modifications, may also inactivate wild-type p53 by inhibiting the transactivation capability of p53 and by targeting p53 protein for degradation.

Although inactivation of p53 may be unnecessary or insufficient for tumor initiation, in many cancers it contributes to malignant progression, metastatic potential, and invasiveness (Hsiao et al., 1994). Tumor development requires adequate and continuously growing blood supply, and the involvement of p53 in angiogenesis explains its participation in later stages of tumor progression . Expression of wild-type p53 has been shown to stimulate inhibitors of angiogenesis (Dameron et al., 1994; Van Meir et al., 1994), and a mutant p53 can participate in the stimulation of the angiogenic vascular endothelial growth factor (VEGF) gene (Kieser et al., 1994). In addition, hypoxia, the usual condition in the center of a tumor, induces p53-dependent apoptosis of cells containing wild-type p53, whereas cells harboring mutant p53 survive allowing tumor expansion (Graeber et al., 1996).

p53 mutations in Li-Fraumeni syndrome

Li-Fraumeni syndrome (LFS) provides a human model of nonfunctional p53. The patients have inherited germ-line p53 mutations, which highly predispose them to multiple primary cancers at an early age (Malkin, 1993). The characteristic neoplasms in this dominantly inherited disorder include breast cancer, sarcoma, and glioma, as well as several more rare tumors arising in childhood (Birch et al., 1994). Families with a germline missense mutation in the DNA-binding domain of p53 possess higher cancer rates at earlier age than the families with other type of p53 mutations (Birch et al., 1998). However, heterozygous germ-line p53 mutations have been found in only 71% of LFS families, and many tumors in these patients show no loss of heterozygosity (Sedlacek et al., 1998), questioning the role of p53 as the sole cause of cancer development (Varley et al., 1997).

p53-/- mice and MEFs

p53-/- mice, contrary to RB deficient mice, are viable but are highly prone to both spontaneous and induced tumors, especially lymphomas (approximately 60 % of tumors) and soft tissue sarcomas (20% of tumors) appearing within first 6 months (Donehower et al., 1992; Harvey et al., 1993a). The tumor incidence in heterozygous p53+/- mice is intermediate compared to normal and homozygous p53 -/- mice, and tumor cells of these mice have often lost the remaining wild-type allele. Interestingly, the tumor spectrum in these animals consists predominantly of osteal and soft tissue sarcomas instead of lymphomas (Harvey et al., 1993b). The difference in tumor spectra between heterozygotes and homozygotes may depend on how easily a particular tissue type loses the wild-type allele or how low levels of p53 can still maintain normal p53 functions. The predisposition of p53-/- mice to lymphomas is also of interest, because the loss of p53 is rarely encountered in lymphomas.

p53-/- MEFs derived from p53-deficient mice show several cellular abnormalities: the cells have shortened halflife, they fail to senescence at high passages, and are able to grow at low density (Harvey et al., 1993c) . In addition, chromosomal abnormalities and aneuploidy appear at early passage giving rise to genomic instability. Cells lacking p53 are also highly tolerant to different genotoxic impulses leading to accumulated, unrepaired DNA lesions (Lowe et al., 1993a).

Loss of p53 may sometimes be sufficient for immortalization of cells, but more often a lack of p53 allows other genetic changes favoring tumor formation, for example oncogene activation, to occur. Thus, depending on the original genetic backround as well as on accumulating of other mutations, a considerably different tumor pattern and incidence may arise.

p53 is a nuclear transcription factor, which activates the transcription of several target genes involved in the regulation of cell growth and apoptosis. Human p53 protein consists of 393 amino acid residues, which can be roughly divided into four structurally and functionally different domains (Fig. 3.). The acidic amino-terminal domain consisting of the first 42 aa is responsible for the transactivating properties of the protein, without which the induction of target genes cannot occur. Amino acids ranging from 102 to 292 form the central sequence-specific DNA-binding domain (Bargonetti et al., 1993; Pavletich et al., 1993), the most common location of p53 mutations. Between these two domains a novel proline-rich domain localized between aminoacids 64 and 92 has been identified (Walker and Levine, 1996). The carboxy-terminal domain displays many functions, and can be further divided into the oligomerization domain (amino acids 324-355) and the basic C-terminus with the last 30 amino acids.

Amino-terminal domain

The N-terminal as well as the C-terminal domain of p53 binds to numerous proteins involved in DNA replication and repair. A list of factors binding to and affecting the transactivation domain includes many general transcription factors, such as TATA box-binding protein (TBP) (Horikoshi et al., 1995) and TBP-associated factors (TAFs) TAF70 and TAF31 (Lu and Levine, 1995; Thut et al., 1995), all of which are subunits of the general transcription factor TFIID. By binding to TBP, wild-type p53 can repress genes lacking a p53-binding site. p62, a component of the dual transcription /repair factor TFIIH, and RP-A, a single stranded DNA-binding protein, also interact with amino-terminus of p53. p53 can thus induce growth arrest not only by activating growth suppressive target genes but also by the repression of transcription by sequestering and binding to TBP, RPA and other proteins involved in replication (Seto et al., 1992; Mack et al., 1993). C-Abl is activated after DNA damage and binds to the N-terminal domain of p53 enhancing its transactivation function (Goga et al., 1995), whereas adenovirus E1B 55 kD protein inhibits p53-mediated apoptosis. Most importantly, the negative regulator of p53, mdm2, binds to the N-terminus of p53, sharing the same amino acids 22 and 23 of p53 with E1B 55 kD protein (Lin et al., 1994). In addition, both N- and C-terminus are subject to diverse phosphorylation modifications that are involved in the regulation of p53; these changes are discussed in the context of p53 regulation.

The proline-rich domain

This domain contains five repeats of the PXXP motif, where P represents proline and X any amino acid (Walker and Levine, 1996). Removal of proline-rich domain does not affect cell cycle arrest but results in an impaired capability of p53 to suppress growth of tumor cells (Walker and Levine, 1996), and impaired apoptotic activity (Venot et al., 1998; Zhu et al., 1999). Interestingly, p53 mutants lacking the proline-rich domain can selectively transactivate certain target genes; natural p21, mdm2, and bax promoters are stimulated while the transactivation of PIG3 is inhibited (Venot et al., 1998). Additionally, this domain seems to be important for transcriptional repression (Venot et al., 1998), which together with selective transactivation is involved in apoptosis mediated by the proline-rich region.

DNA-binding domain

The central sequence-specific DNA-binding domain is an independently folded, Zn²⁺ ion requiring domain. The four conserved regions within the core domain make contact with the major and minor grooves of p53-binding sites, while the less conserved regions serve as a structural element, the ß-sandwich, which helps positioning the residues that interact with DNA (Cho et al., 1994). Proteins defective in DNA binding are incapable of transactivating target genes and transmitting most of the effects of p53. The importance of DNA binding for p53 function is underscored by the fact that p53 residues most often mutated in human cancer, among them Arg248, Arg273, and Arg175, are involved in DNA binding and located in conserved regions (Fig. 3). SV 40 T antigen binds to and blocks the DNA-binding domain, whereas the p53 binding proteins 53BP1 and 53BP2 enhance p53-mediated transcriptional activity (Iwabuchi et al., 1998). Although the tumor suppression function of p53 is highly linked to the core domain, other domains may also be needed for the full suppression of transformation.

The three-dimensional co-crystal structure of the DNA-binding domain (Cho et al., 1994) confirmed the tetrameric nature of p53 molecule. In solution and in wild-type conformation p53 is found as a stable dimer of a dimer, a conformation characteristic but not unique to p53 protein. Four DNA-binding domains together are more effective than one in reaching multiple and sometimes distant p53-binding sites in a target gene promoter (for example in p21 and cyclin G genes) (Zauberman et al., 1995). Tetrameric p53 protein binds to two repeats of a consensus DNA sequence 5´-PuPuPuC(A/T)(T/A)GPyPyPy-3´ (El-Deiry et al., 1992).

The C-terminal domain

A flexible linker connects the DNA-binding domain to tetramerization or oligomerization domain. The structure of this domain has been elucidated by three-dimensional nuclear magnetic resonance spectroscopy (Lee et al., 1994; Clore et al., 1995a; Clore et al., 1995b) revealing a certain homodimerizable motif (ß-sheet-turn-a-helix). Only a small fraction of point mutations in p53 are located in this domain, and some evidence suggests that tumor cells select for an intact oligomerization domain, because tetramerization may be needed for complexing with wild-type p53 and exerting dominant-negative phenotype. Although the role of the oligomerization domain in transformation may be negligible, it is, as well as the activation domain, needed for growth suppression.

The extreme C-terminus is an important and autonomous domain regulating the activation of the whole p53 molecule. It can bind nonspecifically to different forms of DNA-strands (Bakalkin et al., 1994; Lee et al., 1995) and reanneal complementary single strands of DNA and RNA (Brain and Jenkins, 1994; Wu et al., 1995). Normally, the carboxy-terminus acts as an autoinhibitory region, but several factors are able to abolish this negative-regulatory function. p53 function can be activated by C-terminal deletion (Hupp et al., 1992), antibody binding (Hupp et al., 1993; Halazonetis et al., 1993), phosphorylation or other posttranslational modifications (Shaw et al., 1996), or by binding of small C-terminal peptides (Hupp et al., 1995). Monoclonal antibody PAb421 recognising amino acids 370-378 is a well known activator of p53 sequence-specific DNA binding. Besides activating wild-type p53, some of these modifications can reactivate most naturally occurring mutant p53 proteins (reviewed by Selivanova et al., 1998). For example, C-terminal synthetic peptide corresponding to residues 361-382 has been shown to restore the apoptotic and growth suppression functions of at least two common p53 mutations (R248Q and R273H) in human tumor cells (Selivanova et al., 1997). According to the allosteric model, the interaction between the C-terminal regulatory domain and its binding site in p53, probably in the core domain, results in a conformation that is unable to bind DNA. C-terminal modifications are postulated to change the conformation of p53 so that the DNA-binding domain is exposed and sequence-specific binding is allowed. Recently it was demonstrated that the peptide derived from the C-terminal domain of p53 binds to the core domain of mutant p53 (Selivanova et al., 1999). The binding of C-terminal peptide may relieve the inhibition by disrupting the interaction between C-terminus and its binding site by competitive inhibition, and moreover, the peptide may stabilize the DNA-binding domain or establish new DNA contacts (Selivanova et al. 1998).

Short (20-39 nucleotides) single strands of DNA interact with the C-terminus and effectively activate DNA binding of p53, whereas longer and double stranded DNA inhibit this function (Bakalkin et al., 1994; Jayamaran and Prives, 1995). Thus, the role of the C-terminus in damage recognition and DNA repair is significantly based on its ability to 1) bind single stranded DNA, generated for example during replication errors and excision repair processes, 2) catalyze the reassociation of single stranded DNA into the double stranded form, and 3) interact with DNA helicases. Interestingly, antibody pAb421 that activates the DNA-binding function of p53, can at the same time inhibit the nonspecific binding and reannealing activities (Jayamaran and Prives, 1995; Wu et al., 1995). It has been postulated that two different active conformations of p53 may exist, one being activated by an antibody or phosphorylation and leading to enhanced specific but decreased nonspecific DNA binding, and the other being incapable of DNA binding but remaining active for other functions. Thus, many kinds of modulations in the C- and N-terminus can regulate the p53 molecule between inactive and active, perhaps more than one, conformations.

p53 homologues

Quite recently two p53 homologues, p63 and p73, have been identified, sharing considerable homology with the activation, DNA binding, and oligomerization domains of p53 (Kaghad et al., 1997; Yang et al., 1998). However, p73 is not induced by DNA damage, but its overexpression causes growth arrest and apoptosis (Jost et al., 1997). Both p63 and p73 are able to transcriptionally activate p53 target genes depending on which of the multiple isoforms they exist in (Jost et al., 1997; Yang et al, 1998). A tumor suppressor function has been suggested at least for p73, since neuroblastomas often lacking p53 mutations exhibit deletion of the genomic region containing the p73 gene.

Several different cellular and environmental conditions, usually harmful to the cell or organism, cause the accumulation and activation of p53, resulting in growth arrest or apoptosis. Although high p53 levels are detected during development at specific tissues including the central nervous system, p53 protein is otherwise present at a very low level in normal cells. Although the protein is being constantly translated, it is actively and efficiently degraded until needed. The half-life of normal, unactivated p53 protein is less than 30 minutes, whereas posttranslational stabilization increases the time to several hours (Kastan et al., 1991). Additionally, p53 can itself regulate its activity at the translational level. In normal cells translation of p53 mRNA is constitutively inhibited by its 3´ untranslated end. Furthermore, p53 protein can bind to its own mRNA repressing its translation, but upon DNA damage the repression is suppressed (Mosner et al., 1995). The presence of p53 at extremely low levels in normal conditions and its activation upon damaging or otherwise harmful situations suggest that p53 is not needed during the normal cell cycle, but is called into action when problems arise. This mode of function differs from the action of other tumor suppressors, for example RB, which is constantly working throughout the cycle.

Genotoxic stress, for example UV and g-irradiation, and cytotoxic drugs, is a strong inducer of p53 (Fig. 4.). Although the extent and kinetics of p53 accumulation and activation differ somewhat among these stimuli, the common denominator is the presence of DNA strand breaks (Nelson and Kastan, 1994). Besides DNA double-strand breaks, which are directly formed by g-irradiation and indirectly by DNA repair processes after UV radiation and cancer therapy drugs, single strand breaks can also induce p53 (Di Leonardo et al., 1994). Introduction of restriction enzyme nucleases into the nucleus results in an increase in the level of p53 demonstrating that strand breaks are sufficient for activation of p53 (Siegel et al., 1995).

Furthermore, more general stress situations arising from suboptimal growth conditions, such as hypoxia (Graeber et al., 1994), alterations in redox balance (Hainaut and Milner, 1993), heat, starvation, and nucleotide depletion (Linke et al., 1996) induce p53 (Fig. 4.). Since oxidation inhibits and reduction induces DNA binding of p53 (Hainaut and Milner, 1993; Hupp et al., 1993), inactivation of p53 provides a mechanism by which oxygen radicals can promote tumor formation.

In all the above mentioned situations, as a physiological response, p53 is accumulated and activated at the same time, but in other contexts p53 may be accumulated without activation. p53 may complex with viral oncoproteins, for example SV40 T antigen, leading to an increase in the amount of p53 protein but not in activity. On the other hand, E1A and E7 proteins, which complex with pRB instead of p53, cause p53 stabilization and p53-dependent apoptosis (Lowe and Ruley, 1993; Demers et al., 1994). In addition, p53 mutations couple increased p53 protein level with decreased or absent function. In conclusion, all conditions where p53 is accumulated carry an increased risk for transformation, regardless of the mechanism or reason for accumulation (DNA-damage, infection with tumor viruses or mutation of p53).

The p53 tumor suppressor protein limits cell proliferation and inhibits tumor formation by two main mechanisms: by inducing growth arrest and/or apoptosis (Fig. 4.). The organism is protected from accumulating genetical changes by halting the propagation of harmful mutations to daughter cells. This is achieved by either arresting the cell cycle and allowing time for DNA repair, or by committing suicide by programmed cell death. In addition to growth and tumor suppression functions, p53 is involved in development and differentiation processes.

Growth arrest

The best characterized antiproliferative function of p53 is its role in G1-phase growth arrest. g-irradiation of cells rapidly and efficiently leads to p53-dependent G1-phase arrest (Kastan et al., 1992; Kuerbitz et al., 1992). Activation of p53 induces the expression of several target genes, especially p21 Cdk-inhibitor gene (El-Deiry et al., 1993). Upregulation of p21, a potent inhibitor of all Cdks, blocks the catalytic activity of the cyclin-Cdk complex preventing the phosphorylation of many regulatory proteins and leading to inhibition of cell cycle progression (Dulic et al., 1994; El-Deiry et al., 1994). Among the best known consequences is the maintenance of pRB in the underphosphorylated form binding transcription factor E2F-1 (Demers et al., 1994; Slebos et al., 1994). The pRB-E2F complex represses the gene transcription inhibiting the expression of numerous proteins needed in the progression of the S phase. Partly contradictory data have been presented concerning the need of pRB-E2F pathway in p53-mediated growth arrest. In experiments where pRB functions were blocked by viral oncoproteins like the E1A protein of adenovirus, the T antigen of SV40 or E7 of human papillomavirus, p53-dependent growth arrest was inhibited (Demers et al., 1994; Slebos et al., 1994). On the other hand, cells from RB null mice were shown to be capable of undergoing full G1 arrest after g-irradiation. Several aspects should be taken into account here: first, oncoproteins effect many other cell cycle proteins besides pRB, and secondly, RB null cells have intact RB family members p107 and p130 which may partly overtake the functions of pRB. However, in the absence of clear evidence of competing mechanisms, G1-growth arrest is generally concluded to be exerted via the p21-pRB-E2F pathway.

The role of p21 in p53-mediated G1 growth arrest is evidenced by the observation that overexpression of p21 itself leads to arrested growth (Harper et al., 1995). However, p21-/- cells are only partially defective in g-radiation induced growth arrest (Brugarolas et al., 1995; Deng et al., 1995) whereas p53 deficiency totally abolishes it, suggesting that other p53 activated factors, for example GADD45, may be also involved. Besides well characterised G1 growth arrest, p53 has a definite role in mediating G2 growth arrest as well (Agarwal et al., 1995; Stewart et al., 1995).

Even in the absence of damaging stimulus p53 functions as a cell cycle checkpoint controller, since p53-/- MEFs exhibit disturbed cell cycle phase distributions compared to wild-type p53 cells (Harvey et al., 1993). Furthermore, the same cells treated with mitotic spindle inhibitors failed to be arrested in the G2 phase, but continued DNA synthesis inappropriately resulting in aneuploidy (Cross et al., 1995). This suggests that p53 functions as a mitotic spindle checkpoint factor ensuring that replication is not initiated without completion of proper chromosome segregation. Studies with gas1 gene have suggested participation of p53 even in the G0 checkpoint. Gas1 gene, expressing a protein needed to keep cells in G0 state, is not a target gene of p53, but is dependent on wild-type p53 for its function. p53 mutants incapable of transactivation are still able to assist gas1 to arrest cells in G0, giving an example of a p53 checkpoint function devoid of any transactivation property (Del Sal et al., 1995).

Apoptosis

Apoptosis, or programmed cell death, is an efficient means of getting rid of harmful cellular material. Convincing number of experiments have demonstrated a role for p53 in triggering apoptosis in a variety of different conditions. DNA damaging agents, such as ionizing and UV radiation and cytotoxic drugs, c-myc (Wagner et al., 1994) and adenovirus EIA expression (Lowe and Ruley, 1993), and withdrawal of growth factors (Johnson et al., 1993), may result in p53-dependent apoptosis. g-irradiation of mouse thymocytes provides a classic example of p53-induced cell death (Lowe et al., 1993b). Radiated thymocytes from p53 -/- mice do not undergo apoptosis as do their p53 positive counterparts (Lowe et al., 1993b). Since p53-mediated apoptosis can be blocked by treating myeloid cells with interleukin-6 (Levy et al., 1993) or erythroid cells with erythropoietin, p53 may in these circumstances function as sensor of environmental conditions and growth factor supply. Cancer threapy is also taking advatage of apoptotic pathway, since cytotoxic drugs and irradiation therapy are largely based on the wild-type p53-induced apoptosis. It should be remembered, however, that p53-independent apoptosis, such as cell death after glucocorticoid treatment (Clarke et al., 1993), probably occurs as often as p53-dependent cell death.

p53 participates in the induction of apoptosis directly by inducing the target genes bax, insulin-like growth factor-binding protein 3 (IGF-BP3) (Buckbinder et al., 1995), fas/APO1 (Owen-Schaub et al., 1995), and Killer/DR5 (Wu et al., 1997). Bax, an apoptosis promoter, can form a complex with Bcl-2, a protector against apoptosis, and the balance between these two, among other factors, determines whether apoptosis is induced (Miyashita and Reed, 1995). The IGF-BP3 gene is upregulated by p53 in response to genetic insult (Buckbinder et al., 1995). By binding to IGF it inhibits both survival and mitogenic signals representing an antimitogenic function of p53.

In contrast to the growth suppressive function of p53 that seems to be dependent on the transcriptional activity, apoptotic activity has been difficult to uniformly attribute to any certain domain of the protein. Although p53 induces several genes promoting apoptosis, p53-dependent apoptosis after DNA damage has been detected both in the presence of the translational inhibitor cycloheximide (Wagner et al., 1994) and the transcriptional inhibitor actinomycin D (Caelles et al., 1994) suggesting that the apoptotic pathway may be independent of the transcriptional activity of p53. The same conclusion was obtained in HeLa cells transfected with mutant p53_{gln22, ser23}. This mutation which abrogates the transcriptional activity of p53 was still able to cause p53-dependent apoptosis (Haupt et al., 1995; Yonish-Rouach et al., 1995). Accordingly, neither Bax nor fas/APO1 expression were required for p53-dependent apoptosis in vivo (Knudson et al., 1995; Fuchs et al., 1997). In contrast, BRK cells stably expressing E1A and temperature-sensitive p53_val135 underwent apoptosis when p53 was in its wild-type conformation, but not when the mutant status was applied (Sabbatini et al., 1995a). These controversial findings suggest the existence of two separate apoptotic functions of p53, one depending on the activation of p53 target genes, and the other using direct protein signaling, depending on the cell type or experimental situation. Repression of certain genes may offer one mechanism for transcription-independent apoptosis, since bcl-2 and adenovirus E1B 19 kD protein blocking the p53-dependent apoptosis do not interfere with transcriptional activity but suppress the repression function (Shen and Shenk, 1994; Sabbatini et al., 1995b). This is supported by observation that lack of proline-rich domain of p53 that results in defective apoptosis also impairs the repression function of p53 (Venot et al., 1998). Finally, the large number of proteins interacting with and binding to p53 give rise to numerous possible mechanisms by which p53 may induce apoptosis independent of transactivation.

Growth arrest or apoptosis?

DNA-damaging stimuli and stressful environmental conditions lead to either growth arrest of cells or apoptosis. Several factors determine which way the cells follow. One of the most important factors is the cell type: in mortal primary fibroblasts DNA-damage often leads to transient growth arrest, whereas immortal and hematopoietic cells often react by programmed cell death (Midgley et al., 1995; Haupt et al., 1996).

In addition to DNA-damaging situation, limited availability of growth factors or activation of cellular or viral oncogenes favor the induction of apoptosis rather than growth arrest, in order to eliminate cells with unstable genomes. Different viral proteins have diverse and partially controversial effects on p53 and cell survival. Expression of adenovirus E1A protein in murine fibroblasts stabilize and activate p53 resulting in apoptosis (Debbas and White, 1993). On the other hand, 55 kD E1B protein binds to the N-terminus of p53 and inhibits its transcriptional activity, whereas 19 kD E1B protein acts like bcl-2 blocking the apoptosis downstream of p53 activation (Debbas and White, 1993). Human papillomavirus E7 protein induces p53-dependent apoptosis, but E6 binds to p53 and targets it for proteosome degradation (White et al., 1994). Both apoptosis promoting oncoproteins E1A and E7 share the ability to bind to and inactivate pRB, not p53, leading to unregulated functions and increasing amounts of E2F-1. Thus, in the presence of intact p53 but in the absence of functional pRB, the apoptotic pathway rather than growth arrest is triggered by these viral oncoproteins.

p53 in DNA repair

As the role of p53 in growth arrest and apoptosis is widely accepted, the involvement of p53 in DNA repair is far from clear. Studies for and against a role of p53 in DNA replication and repair exist. Interaction with many replication and repair proteins, for example RP-A (Dutta et al., 1993), and components of the dual transcription-repair factor TFIIH, including DNA helicases ERCC2 and ERCC3 (Wang et al., 1995; Leveillard et al., 1996), has been the first line of evidence for p53 in these functions. The binding of p53 inhibits the helicase activity of these proteins, as has been shown also with other cellular and viral helicases (Wang et al., 1996). Indirectly, upon stimulation of p21, which complexes with proliferating cell nuclear antigen (PCNA), p53 is involved in the inhibition of DNA replication after DNA damage without interfering with the repair processes (Shivji et al., 1994; Li et al., 1996).

Although in some studies a correlation has been found between p53 deficiency and reduced repair, in others this has not been the case. UV-radiated Li-Fraumeni cells carrying homozygous deletion of p53 exhibited disturbed removal of pyrimidine dimers and photoproducts but normal transcription-coupled repair (Ford and Hanawalt, 1995; Ford and Hanawalt, 1997). However, point mutations accumulated equally (Nishino et al., 1995; Sands et al., 1995), and the rate of NER was identical in both normal and p53-/- mouse cells (Ishizaki et al., 1994) giving evidence against direct p53 functioning in DNA repair. Nevertheless, the loss of p53 reduced the rate and efficiency of nucleotide excision repair in human cells in several experimental conditions (Smith et al., 1995; Wang et al., 1995). One can not either ignore the fact that p53 binds efficiently to damaged DNA, both mismatched and irradiated DNA (Lee et al., 1995; Reed et al., 1995), and that 3´-5´ exonuclease activity of p53 may be used in DNA recombination, replication or repair processes (Mummenbrauer et al., 1996).

Development

Although p53 -/- mice first appeared to develop normally (Donehower et al., 1992), more careful examination of the phenotype revealed developmental disturbances, particularly defects in neural tube closure. For example, at the age of 13.5 days 16% of p53 -/- embryos, all female, showed markable anencephaly (Armstrong et al., 1995; Sah et al., 1995). Additionally, a small fraction of p53 knockout embryos died prematurely, and the embryos of p53-/- pregnant mice were more vulnerable to diverse teratogenes compared to normal mice (Nicol et al., 1995). Mdm2 -/- mice are embryonically lethal, but the absence of p53 in double knockout mice is able to rescue the embryos (Jones et al., 1995; Montes de Oca Luna et al., 1995). This suggests that in the presence of p53 mdm2 is needed during embryogenesis to negatively regulate p53. Thus, contrary to what was originally believed, p53, in co-operation with mdm2, is not dispensable in early development. p53 is also involved in certain differentiation processes, like B-cell maturation (Aloni-Grinstein et al., 1993) and spermatogenesis (Schwartz et al., 1995).

The main function of p53 is the transcriptional activation of its target genes. Numerous genes have been presented as candidates for effector genes of p53, but not all of them are able to fulfill the criteria. Even if a p53 binding site is found on a gene or promoter, it should be recognized by wild-type but not mutant p53, and basal transcription of a gene should be increased p53-dependently. Besides containing a p53 response element, a gene should be induced upon cellular stress in a p53 dependent manner. Over the years, some candidates have consolidated their positions as p53-response genes and have shown particular importance in the physiological functions of p53.

p21^WAF1/Cip1

Besides being the first p53 target gene discovered, p21 is among the most important effector genes. As a universal Cdk-inhibitor, p21 has a central role in p53-induced growth arrest. p21 functions are discussed thoroughly in a separate chapter.

mdm2

Originally mdm2 (murine double minute gene 2) was described as one of the genes amplified in the double minute chromosomes in spontaneously transformed 3T3 cells (Cahilly-Snyder et al., 1987), and later it was found to possess oncogenic potential (Fakharzadeh et al., 1991). Mouse mdm2 gene contains two promoters and numerous exons spread over a large area. One promoter is located upstream of the gene and used for constitutive expression, while the other is present in the first intron and regulated by p53 through two p53 binding sites (Juven et al., 1993; Wu et al., 1993; Barak et al., 1994; Zauberman et al., 1995). Basal expression of mdm2 from its constitutive promoter P1 is thought to be responsible for maintaining low levels of p53, while p53-inducible mdm2 promoter P2 is not activated after DNA damage until interaction between mdm2 and p53 has been disrupted and activation of the target genes has occured. In order to give p53 time to act before it is attenuated by mdm2, transcriptional activation of mdm2 should occur later than activation of other target genes, or, alternatively, the complex formation between p53 and mdm2 can be retarded.

Mdm2 oncogene encodes several proteins, and the largest 90 kD protein binds to and inactivates p53 (Momand et al., 1992; Haupt et al., 1997; Kubbutat et al., 1997). Multiple other mdm2 proteins have also been detected, and these proteins arise through alternative splicing, proteolytic processing, or posttranslational modifications (Olson et al., 1993). p76, which is unable to bind p53 but binds ARF, is expressed and induced by UV radiation p53 independently (Saucedo et al., 1999). UV radiation has been shown to induce mdm2 in a p53 and UV-dose dependent manner: low UV doses induce the gene rapidly, while at high doses a delay is observed between mdm2 induction and p53 stabilization (Perry et al., 1993).

The mdm2 gene is amplified in about 20% of soft tissue sarcomas (Oliner et al., 1992; Leach et al., 1993) and quite frequently in esophageal tumors, astrocytomas and osteosarcomas (reviewed by Momand and Zambetti, 1997). In addition to gene amplification high levels of mdm2 may result from enhanced translation or other modifications (Landers et al., 1994). Although p53 is mutated in half of human cancers, p53 mutation is rarely found together with mdm2 amplification (Florenes et al., 1994a). Instead of p53 mutation, exceptionally high levels of wild-type p53 are observed in some cancers. Transformed epithelial cells expressing high basal level of wild-type p53 were found to overexpress mdm2 without gene amplification, and with no increase in transcription of other p53 effectors (Blaydes et al., 1997). Additionally, only a small proportion of p53 was found in a complex with mdm2, so that the amount of functional p53 was not totally abolished but was lowered to a reasonable, more physiological level (Blaydes et al., 1997). This suggests that overexpression of mdm2 provides a mechanism by which high wild-type p53 levels can be tolerated without growth arrest or the apoptosis of cells.

The main contribution of mdm2 overexpression to cancer formation is believed to be through elimination of the tumor suppressor function of p53. However, the N-terminal domain of mdm2 is able to activate E2F-1 transcription factor (Martin et al., 1995), and to bind to and inactivate pRB (Xiao et al., 1995). These functions may explain how mdm2 promotes transformation also in cells lacking p53 (Lundgren et al., 1997).

GADD45

GADD45 is the best characterized member of the growth arrest and DNA damage inducible gene (GADD) family. These genes are induced by DNA-damaging agents and some growth arrest conditions, for example nutrient depletion or hypoxia (Fornace et al., 1989). GADD45 is a nuclear protein, the levels of which oscillate slightly during the cell cycle being highest in G1 and lowest during the S phase (Kearsey et al., 1995). GADD45 can be transcriptionally activated by p53 via p53 consensus binding site at its third intron (Hollander et al., 1993). Although induction of GADD45 upon g-irradiation is strictly dependent on p53 (Kastan et al., 1992), p53 independent induction of GADD45 is often seen after base-damaging agents, such as UV radiation (Kearsey et al., 1995). Even if p53 may not be necessarily required for the induction of GADD45, disruption of p53 function may reduce the extent of induction (Zhan et al., 1996). An increased level of GADD45, either due to DNA damage or overexpression of the protein, leads to G1 growth arrest of cells (Zhan et al., 1994). A role for GADD45 in DNA repair and inhibition of apoptosis has been suggested by experiments where antisense GADD45 expression decreased DNA repair and sensitized cells to UV- and cisplatin-induced apoptosis, the latter function being opposite to that of p53 (Smith et al., 1996). Moreover, in some experiments GADD45 has been shown to interact with PCNA, and, like p21, inhibit PCNA-dependent replication and thus facilitate nucleotide excision repair (Smith et al., 1994; Hall et al., 1995). However, other studies have not been able to demonstrate the effect of GADD45 on NER (Kearsey et al., 1995).

bax

Bax is a target gene of p53 involved in apoptotic function of p53. Bax forms heterodimeric complexes with bcl-2 and by dominant-negative effect antagonizes its ability to block apoptosis. Interestingly enough, also Bcl-2 seems to be under the control of p53, because p53 can downregulate bcl-2 by repressing its transcription (Haldar et al., 1994; Miyashita et al., 1994).

Others

Cyclin G encodes a cyclin of unknown function (Okamoto and Beach, 1994), which is induced in a p53 dependent manner after DNA damage (Miyashita and Reed, 1995); no Cdk has been yet found to associate with it. The regulation of PCNA by p53 has been questioned, but it seems that transactivation of PCNA is restricted to low concentrations of p53 (Shivakumar et al., 1995; Morris et al., 1996). Additionally, numerous other p53 response genes have been suggested, but the real in vivo importance of these target gene candidates needs to be clarified.

Transcriptional repression by p53

Besides being a transcriptional activator, p53 may function as a transcriptional repressor as well, providing yet another mechanism to influence cellular functions. Genes lacking p53-binding site but containing the TATA box can be subject to inhibition by p53 (Mack et al., 1993). c-fos, c-jun, RB, bcl-2, IL-6 and heat shock protein (hsp) 70 are examples of genes repressed by p53. The repressor function of p53 is believed to be especially important in suppression of tumorigenesis and induction of apoptosis.

The p21^WAF1/CIP1 cyclin kinase inhibitor cDNA was cloned independently by four different groups and thus was given several different names characterizing its functions. Using a yeast-two hybrid system, p21 was reported to bind to Cdk and was named CIP1 (Cdk inhibiting protein 1) (Harper et al., 1993), and, at the same time, it was identified as a gene product activated by p53, called WAF1 (wild-type p53-activated factor 1) (El-Deiry et al., 1993). Later, when DNA synthesis inhibitors were looked for from senescent cells, p21 was cloned as SDI1 (senescent cell-derived inhibitor 1) (Noda et al., 1994). Finally, p21 was identified as MDA-6, melanoma differentiation associated protein, because of its increased expression in melanoma cells after induced differentiation (Jiang et al., 1994).

Growth arrest

The main effect of p21 seems to be the inhibition of cell proliferation. As a common cyclin-kinase inhibitor, p21 effectively inhibits the kinase activity of Cdk2, Cdk3, Cdk4 and Cdk6, less so Cdk5 and not at all Cdk7 or Cdc2 (Harper et al., 1995). As a result, many regulatory proteins needed for progression of the cell cycle remain unphosphorylated and the cell cycle stops. Originally, p21 was assumed to induce growth arrest only at the G1/S-phase transition, but recently it has been shown to mediate G2 arrest as well (Dulic et al., 1998; Niculescu et al., 1998). p21 may induce growth arrest by at least two different mechanisms: by inhibiting the kinase activity of Cdks and by preventing the actions of PCNA. Two different domains of p21 are responsible for these inhibitory functions (Chen et al., 1995; Luo et al., 1995). The N-terminal part of p21 is required for the Cdk inhibition, while the carboxy-terminal domain interacts with PCNA and prevents the interaction of PCNA with other components of the polymerase assembly. The association with PCNA is a unique feature of p21 that distinguishes it from other CKIs. As an auxillary factor for DNA polymerases e and d, PCNA is essential for both DNA replication and DNA repair. Although inhibiting the PCNA-dependent replication and mismatch repair in vitro (Flores-Rozas et al., 1994; Li et al., 1994; Waga et al., 1994), p21 appears not to inhibit the function of PCNA in nucleotide excision repair (Shivji et al., 1994; Li et al., 1996).

In normal human fibroblasts, p21 is found in quaternary complexes with cyclin, Cdk and PCNA (reviewed by Kelman, 1997). Cdk in this complex may be inactive or active depending on the amount and stoichiometry of p21 protein. At low or intermediate concentrations p21 seems to function as an assembly factor enforcing the assembly of cyclin D and Cdk4 even by 35-fold and targeting the cyclin/Cdk complex to the nucleus, whereas at high concentrations p21 works as an inhibitor of kinase activity (LaBaer et al., 1997). In a recent study, the antiproliferative functions of p21, inhibition of cyclin/Cdk complexes and inhibition of PCNA, were shown to be independent of each other, the latter function alone being sufficient to block cell cycle progression at the G1/S and G2/M transitions (Cayrol et al., 1998). Interaction of p21 with GADD45 damage response protein is probably also involved in DNA damage checkpoints (Kearsey et al., 1995).

The role of p21 in DNA damage checkpoints has been studied with mice and human cells with homozygous deletion of p21 gene. Gamma-irradiated p21-/- MEFs were partially defective in G1 checkpoint control and in response to double-stranded DNA lesions (Brugarolas et al., 1995; Deng et al., 1995). However, in human colon carcinoma cells HCT116, p21 was absolutely required for p53-mediated G1 growth arrest by ionizing irradiation (Waldman et al., 1995).

Quite opposite to growth arrest, mitogenic stimuli have ben observed to induce p21 (Macleod et al., 1995). Enhanced expression of antiproliferative p21 may be perhaps seen as a compensatory mechanism to prevent the overshoots of mitogenic activation.

Terminal differentiation and senescence

p21 is often induced in the process of terminal differentiation during which cells need to exit from the cell cycle. For example, skeletal muscle cell differentiation correlates with MyoD-induced p21 expression (Halevy et al., 1995). Ectopic expression of p21 has been shown to promote differentiation of hematopoietic cells, such as megakaryoblastic leukemia cells CMK (Matsumura et al., 1997) and myelomonocytic cell line U937 by vitamin D₃ (Liu et al., 1996). Accordingly, several growth factors are able to induce p21 expression; specific examples of the regulation of p21 by some of these are mentioned below.

In addition to being discovered to be an inhibitor of DNA synthesis in senescent cells, disruption of p21 function in normal fibroblasts has been observed to lead to a bypass of senescence and an extended life span of cells (Brown et al., 1997).

Apoptosis and tumorigenesis

Although p21 has been considered inert in apoptosis, a continuous line of evidence suggests that p21 may protect cells from apoptosis (Polyak et al., 1996; Gorospe et al., 1997; Bissonnette and Hunting, 1998), while in some studies p21 has been shown to promote it (Duttaroy et al., 1997). In contrast to tumor suppressive Cdk-inhibitor p16 and p14^ARF, p21 appears not to have any role in tumorigenesis. Firstly, p21 mutations are almost never found in human cancers and secondly, p21 knockout mice are not prone to tumors (Deng et al., 1995).

Regulation of p21 protein involves both transcriptional and posttranscriptional mechanisms. Regulation at the transcriptional level may be either p53-dependent or -independent.

p53-dependent induction of p21 transcription

As an effector gene of p53, p21 promoter contains two p53-responsive elements, whose sequences are highly conserved among mouse, rat and human (El-Deiry et al., 1995); one of these sites may suffice for p53 induction (Poluha et al., 1997; Wu and Schonthal, 1997). Several different conditions and factors are able to induce the transcription of p21 in a p53 dependent manner, but DNA-damaging agents, g-irradiation in particular, often induce p21 p53-dependently (Macleod et al., 1995). Accordinly, spindle disrupter nocodazole, differentiation factor Neu, and ribonucleotide synthesis inhibitor PALA, to list just a few, induce p21 by activating or stabilizing p53 (reviewed by Gartel and Tyner, 1999). Quite controversial findings have been reported about the dependency of p53 on the activation of p21 by UV radiation. Some groups have found p53-independent p21 induction after UV-treatment (Loignon et al., 1997), while others have shown a lack of p21 expression in the absence of p53 (Gorospe et al., 1998).

Oncogenic Ras causes growth arrest in primary fibroblasts and leads to accumulation of p53, p21 and p16 (Serrano et al., 1997), but p21 induction by Ras may occur also in a p53 independent manner (Kivinen et al., 1999). Induction of p21 by Raf, a downstream effector of Ras, has been shown to occur both p53-dependently (Lloyd et al., 1997) and independently (Sewing et al., 1997; Woods et al., 1997).

p53-independent regulation of p21 expression

Differentiation promoting agents often induce p21 by p53-independent mechanisms by inducing the binding of different transcription factors to specific elements in p21 promoter. Human p21 promoter contains six Sp1-binding sites located within 120 bp upstream of transcription initiation site, and many p21 activating agents, for example tumor suppressor BRCA-1 (Somasundaram et al., 1997), utilize these binding sites. Accordingly, p21 activation by TGF-ß is mediated by interaction of Sp1 with Smad proteins (Li et al., 1998; Moustakas and Kardassis, 1998). On the other hand, growth arrest and upregulation of p21 induced by EGF and IL-g occur by STAT1 via several STAT-binding sites (Chin et al., 1996). In addition, specific binding sites for transcription factors E2Fs and AP2 exist in p21 promoter (Gartel and Tyner, 1999).

Although p21 regulation occurs mainly at the transcriptional level, posttranscriptional mechanisms are also possible. Both p21 mRNA (Gorospe et al., 1998) and protein (Timchenko et al., 1996) can be stabilized without an increase in its transcription. Posttranscriptional events have been suggested to possess a significant role in p21 regulation particularly after genotoxic stress (Butz et al., 1998).

Malignant melanoma is one of the most common tumor types in western countries; only cancers of breast, prostate, colon, and lung occur more frequently than melanoma. Although epidemiological studies have strongly demonstrated that exposure to sunlight is the major environmental cause of melanoma, the exact molecular targets of UV radiation remain to be identified. The genetic loci 6q and 9p21 are often lost in cutaneous melanomas- the melanoma susceptibility locus at 9p21, including for example p16 gene locus, is absent in over half of malignant melanomas.

During the last years p16 cyclin kinase inhibitor has established its role as a tumor suppressor in familial melanoma as well as in many other cancers. The pedigrees of familial melanoma with several affected individuals in several generations are rare, but over 50% of these families carry a germline p16 mutation (Hussussian et al., 1994). Family members with hereditary susceptibility to melanoma, comprising 5-10% of all melanoma patients, also have a positive family history with at least one affected relative. Germline p16 mutations are observed in about 20% of these patients (FitzGerald et al., 1996). Germline mutations are typically point mutations or small deletions that are located in the exon 1a thus affecting only p16 without interfering with p19^ARF (Gruis et al., 1995; Holland et al., 1995; Walker et al., 1995).

Two familial melanoma kindreds have been found to carry Cdk4 point mutations that make Cdk4 insensitive to the inhibitory actions of p16 (Zuo et al., 1996). Additionally, 92% of sporadic melanoma cell lines have been found to harbor inactivated p16 or Cdk4 proteins, stressing the importance of p16 inactivation in the development of melanoma (Castellano et al., 1997).

The role of p21 in melanomas is theoretically interesting for two reasons: p21 is located in the same area of chromosome 6 which often contains changes in melanoma, and p21 was originally cloned also as a melanoma differentiation associated gene (mda-6). Despite these connections, p21 mutations are practically never found in melanomas. In contrast to p53, p21 expression often decreases with tumor progression (Maelandsmo et al., 1996).