Ординатура / Офтальмология / Английские материалы / Ocular Oncology_Albert, Polans_2003

results, based on unilaterality or bilaterality, sex, age at diagnosis (bilateral average at 15 months and unilateral average at 24 months), family history, and published reports [10], the author suggested that two genetic mutations are required for the development of a retinoblastoma (two-mutation or two-hit hypothesis). Each of those alterations was calculated to occur at a rate of approximately 2 6 10 7 per year. Since a great deal of cell divisions take place in the developing retina, the chance of one mutation occurring is very good. In the dominantly hereditary forms, a first mutation occurring in the germline (either transmitted by one of the parents or occurring de novo) represents the first hit; it will be carried in all developing cells and can be transmitted to descendants. RNB arises from a retinoblast hit by a second somatic mutation. Because this second mutation is likely to occur in more than one embryonic retinal cell that already bears a mutation, most individuals carrying a germline mutation will develop ‘‘early’’ multifocal and/or bilateral disease (in his publication, Knudson took an average of m ¼ 3 for the number of tumors), although 12% develop a unilateral, unifocal lesion [11]. If no second somatic mutation occurs in the developing retina, no tumor will develop. In patients with nonhereditary RNB, the tumor develops from on single retinoblast hit by two successive somatic mutations. Because the occurrence of two consecutive somatic mutations in more than one retinal cell is very unlikely, these patients develop one ‘‘late’’ unifocal unilateral tumor. This two-hit hypothesis does not imply that those two mutations are sufficient for the development of RNB; it remains controversial whether mutations of other genes are also needed, representing the rate-limiting step in tumorigenesis.

Comings, in 1973, subsequently suggested that in an autosomal dominantly inherited tumor such as RNB, the two successive mutations inactivate both alleles of a specific regulatory gene and a tumor suppressor gene and that each mutation is recessive at the cellular level [12]. Benedict et al. later proposed a similar model [13].

In 1983, Benedict et al. and Cavanee at al. obtained the evidence that the two mutations required for tumor formation affected two alleles of the same gene [14]. Thus homozygous loss of function of the RB1 gene initiates retinoblastoma tumor formation. Because the wild-type allele prevents tumor formation, the RB1 gene was called a tumor suppressor gene. The RB1 locus was then mapped to the long arm of chromosome 13 [13q14] by linkage and deletion analysis of the RNB family, in a region close to the gene coding for the esterase D (ESD) [3–6]. The RB1 gene was subsequently cloned in 1986 and characterized [15–18].

III.RB1 GENE STRUCTURE AND PROTEIN STRUCTURE AND FUNCTION

RB1 was the first gene described as a tumor suppressor, and this opened the way to a better understanding of cancer genesis. The structure and function of this gene in the cell cycle and in differentiation are coming to be more understood and are the subjects of extensive reviews [19–26].

A.The RB1 Gene

The RB1 gene is located on chromosome 13 at band q14. It is a large gene that spreads over 180 kb, contains 27 exons that range in size from 31–1873 bp and

Hypophosphorylated pRB binds to and regulates the function of other proteins involved in cell proliferation, including BRG-1, c-Abl, Mdm2, and MCM7. It also plays a unique role in differentiation pathways by leading the precursor cell to exit the cell cycle and start terminal differentiation (for review, see Refs. 25 and 26). Its role in myogenesis, hematopoiesis, and neurogenesis has been described, and different transcription factors seem to be regulated by pRb: CAAAT/enhancer binding proteins C-EBP, the HMG family member HBP1, and the basic helix-loop- helix (bHLH) transcription factors MyoD and Id-2.

2.Structure of pRb (Fig. 1)

pRB is composed of different structural domains that are important for its function. A better knowledge of the relation between structure and function helps us to understand the consequences of mutations and to build strategies for better screening.

Two central domains, the A and B boxes, highly conserved among species and separated by a less conserved spacer region, are required to form the central pocket. This A/B pocket is fundamental for pRB function through its ability to interact with different types of proteins. Two types of bindings have been described: interaction involving an LXCXE motif (Lys-X-Cys-X-Glu), such as interaction with endogenous proteins like HDAC 1 and 2 or viral oncoproteins (simian virus 40 large T antigen, human papillomavirus E7, and adenovirus E1A), and interaction with a non-LXCXE motif, such as interaction with the E2F family of transcription factors, C/EBP and BRG1. In 1998, Lee et al. elucidated the crystal structure of the LXCXE binding site in the pocket and discovered that it consists of five highly conserved amino acid residues separated in the linear peptide sequence [36]. The LXCXE binding site is located inside the B box in a well-conserved sequence, but the A box is essential for proper folding and the formation of an active pocket. Phosphorylation of this site disrupts the interface of the A/B boxes and could be a possible mechanism of reversible inactivation of pRB during the cell cycle.

Although no single ‘‘hot spot’’ for a preferential mutation site has been identified in the RB1 gene, most of the mutations giving rise to a stable mutant pRB affect the protein-binding function of this pocket, pointing out its fundamental role in cell cycle control [37]. Functional assays for the pocket binding activity have even been proposed as a screening method for RB1 mutation [38].

The carboxy-terminal region also plays an important role for tumor suppression, but it is less structured than the A/B pocket. It contains several functional sites, including a nuclear localizing signal, cyclin binding motifs, and seven consensus cyclin-dependent kinase (CDK) phosphorylation sites, and a second E2F binding site. The C-terminal region can also bind the oncoproteins Mdm2 and c-Abl. It can interact with the A/B pocket and regulate its activity. This interaction is strengthened by phosphorylation of the C-terminal residues, which disrupt the interaction A/B pocket/LXCXE proteins.

The N-terminal region is the least-characterized portion. It is also able to interact with the A/B pocket and regulates its activity, probably in promoting a stable active pRB conformational state. It also contains six consensus CDK phosphorylation sites, which may play a role in regulating pRB in the cell cycle and binding sites for different proteins such as MCM7.

IV. RETINOBLASTOMA TUMOR AND MUTATIONS OF RB1

Some 55–65% of patients affected by retinoblastoma are sporadic cases with unilateral lesions, and the first hit involves a somatic mutation. Those sporadic cases are at no risk for transmitting the disease to their offspring. At the opposite extreme are the other 35–45% of hereditary cases of RNB either by transmission of a germline mutation from an affected parent (10% of the patients present a family history) or by a new germline mutation that can be transmitted to offspring. The paternal allele is most often the carrier of the de novo mutation [39,40]. Indeed, 85% of new germline mutations affect the father’s allele [41], but no paternal age effect has been found [40]. This paternal preference is very helpful in predicting the copy of the gene that will be the carrier of the first mutation. Its explanation could be found in the greater number of cell divisions occurring during spermatogenesis than during oogenesis, since mutations usually arise during DNA replication. On the other hand, a somatic mutation does not disclose any maternal or paternal preference for the first inactivated allele.

The hereditary condition is segregated as a Mendelian autosomal dominant trait with 90% penetrance. Among those genetic cases, patients typically develop multifocal and bilateral lesions, although 10–12% develop unilateral tumor [11]. Patients carrying a germline mutation will also be more susceptible to the development of secondary malignancies, especially sarcomas, and 5–7% present midline intracranial primitive neuroectodermal tumor (trilateral retinoblastoma) [42]. Thus, identification of a germline mutation is essential for better follow-up and improved vision and survival prognosis (see Fig. 3 to report the actual risk data for genetic counseling).

A.Mutation Detection

Whenever possible, analysis of tumor cells will give the best information, since the mutation is always present there and homozygous in 70% of cases [14]. If tumor samples are not available for DNA analysis (no fresh tissue harvested from an enucleated eye or local treatment chosen), constitutional cells, most often leukocytes from peripheral blood, will be studied for mutation analysis in clinical laboratories.

Different tests that vary in applicability and cost have been used to study mutations in RNB. Genetic linkage analysis has been developed but it can be applied only to RNB families, necessitating the availability of multiple informative family members (representing only 10% of RNB cases) and are not suitable for de novo mutations, which represent most of the hereditary cases [43,44]. Cytogenetic testing looking for chromosomal anomalies in 13q14 detects only very large chromosomal anomalies, and its low resolution allows the detection of only 7–8% of the patients with bilateral retinoblastoma and 1–4.9% in sporadic unilateral retinoblastoma [45]. Patients carrying a large cytogenetic deletion may be affected by a 13q deletion syndrome, which is associated with developmental delay, mental retardation, and facial dysmorphism. With the isolation and sequencing of RB1, it became possible to analyze the coding region of the gene directly. More sophisticated techniques have thus been developed to allow a greater resolution for the identification of small deletions, insertions, or point mutations, which constitute most of the germline mutations: DNA fragment analysis techniques using Southern blot hybridization

Figure 3 Evaluation of the genetic risk of retinoblastoma (RNB).

with cDNA and genomic clones, which allow only about 15% of mutations in hereditary retinoblastoma to be identified [46,47]; ribonuclease protection assay and the polymerase chain reaction (PCR) [48]; direct genomic sequencing [49,50]; high-resolution gel electrophoresis and multiplex PCR [51]; single-strand conformation polymorphism (SSCP) analysis [52]; and exon-by-exon PCR-SSCP analysis [53,54].

These techniques can be costly and time-consuming in routine clinical testing. To assess this question, Noorani et al. have compared the cost of molecular screening to conventional repeated ophthalmological examinations for a prototypical family of a proband and seven at-risk relatives and demonstrated the value of genetic screening [55]. They found that genetic testing (using fragment analysis by quantitative multiplex PCR and if negative, sequencing of the promoter region and the 27 exons until a mutation was found) was four times less expensive than conventional screening [56]. They suggest complete retinal examination without anesthesia at birth and every 6 weeks until 3 months of age, thus three exams, then examination under general anesthesia at 5, 7, 9, 12, and 16 months and thereafter every 6 or 12 months up to at least 6 years of age, depending of the actual risk of developing a new tumor. The molecular approach is also valuable, since it spares stressful ophthalmological examinations to unaffected siblings and increases vigilance for an appropriate treatment and follow-up for the tested carriers. It also provides more information for a better understanding of the pathogenesis of retinoblastoma.

B.Mosaiscism—To Be Taken into Account in Genetic Analysis

According to Lohmann et al., however, genetic testing fails to detect about 17–20% of the mutations [57]. In those cases, it is possible that the mutation was missed, but they could also reflect a mosaicism that would not be detected in the peripheral blood.

The mosaic could be either the first affected patient in a family or one of the parents. Mosaicism occurs when a mutation in RB1 arises at some point during embryogenesis, and the time point at which the mutation occurs determines the number and type of cells that will carry the defect. The possibility of this mosaicism should be taken into account during genetic counseling, because it could cause significant errors in predicting the actual risk: a patient with retinoblastoma could be a mosaic, bearing the mutation in leukocytes but not in the germline, and thus could be improperly considered to be at risk of transmitting the trait to descendants. At the opposite extreme, the mutation could be present in the germline and in some somatic cells without being detectable in the leukocytes. This could then give false reassurance about the genetic risk for the offspring and also about the occurrence of multifocal and/or bilateral disease for the patient [58]. Sippel et al. studied 156 documented RBN families to evaluate the incidence of mosaicism among them [59]. The investigators were able to determine a mosaicism in 10% of those families, either in the proband or in one of his or her parents. This percentage might be higher since, for some cases, the authors could not gather all the elements characterizing a mosaicism. Thus mosaicism should be taken into account in providing genetic counseling based on DNA analysis, and Sippel et al. propose that genetic tests of germline DNA be performed when feasible for an accurate evaluation of the actual risk.

C.Nature of the First Hit

RNB results from inactivation of both copies of the RB1 tumor suppressor gene. The first allele is usually inactivated by an intragenic mutation (either in the germline or in a somatic cell). The inactivation of the second copy can occur by any mechanisms involved in the first mutation but most commonly (65–70% of cases) is lost by mechanisms involving chromosomal interaction leading to loss of heterozygosity (LOH) at the RB1 locus [14].

Oncogenic mutations leading to at least the inactivation of one of the RB1 alleles have been well documented in the literature. The majority of initial mutations found are predicted to create premature stop codons as a result of nonsense or frameshift mutations due to points mutations, deletions, and/or insertions or to affect the A/B pocket structure. Lohmann et al. studied 119 patients with hereditary RNB; among 83% causal mutations identified in peripheral blood, the authors found 15% of large deletions, 26% of short-length mutations, and 42% of single-base substitutions [57]. No ‘‘hot spot’’ was found on the RB1 gene for a preferential location for single base-pair substitution, but mutations were unequally distributed along the gene. Blanquet et al. found that exons 3, 8, 18, and 19 were preferential targets for mutations among 232 patients affected by hereditary and sporadic RNB [60]. Nevertheless, no mutation was identified in the last three exons, 25, 26, and 27.

This observation led the authors to suggest that mutations in the 30 end might not be oncogenic.

Harbour performed a metanalysis from 19 international reports published from January 1987 to June 1997, comprising 192 patients [61]. The author studied the distribution of reported germline mutations on the RB1 gene for retinoblastoma patients and found that 43% of reported mutations were nonsense mutations and 35% frameshift mutations. These two types of mutations are the most deleterious for protein function, since they usually induce premature stop codons and thus give rise to a truncated protein. Of these mutations, 12% were found in a conserved 50 or 30 splice site of an intron and probably led to aberrant splicing and deletion of one or more exons and thus to a truncated protein. At the opposite extreme, 6% were missense mutations and 3% small in-frame deletions. Those mutations do not result in a truncated protein but rather affect amino acids that are critical for pRb function (amino acid substitution that will disrupt the A/B pocket interaction). Some 2% of the mutations were located in the promoter region and were all associated with lowpenetrance RNB, which represents 4% of the cases included in this metanalysis. In 59%, the mutation was a base substitution. These are mostly recurrent C-to-T transitions located at CGA arginine codons within the open reading frame, which gives rise to a stop codon. This reflects the high mutability of 50-methylated cytosines CpG in CpG dinucleotides through deamination [62]. Local quasirepeat sequences creating a misalignment can also associate to CpG deamination to originate a point mutation [63]. Some 32% of mutations were due to small deletions and 9% to small insertions. The deleted or inserted sequence usually ranges from 39 to þ 55 bp [64]. Only minimal variability was found among publications from different countries. Lohmann, in 1999, reported similar results [64]. The author also created a useful database on a website listing all the oncogenic mutations reported on the RB1 gene (http://www.dlohmann.de/Rb/).

D.Second Hit and Loss of Heterozygosity

Loss of the homologous normal allele is the most common somatic event that leads to the inactivation of the second wild-type allele of the RB1 gene (second hit) in a cell that is already heterozygous after inactivation by a first mutation in the other allele. This results in homozygosity or hemizygosity for the abnormal allele. Loss of heterozygosity (LOH) was first described in retinoblatoma by Cavenee et al. in 1983 in cultured RNB cells [14] and confirmed by Dryja et al. on RNB family studies [65]. This mechanism plays an important role in the expression of recessive mutations and inactivation of other tumor suppressor genes [66].

LOH represents 50–70% of all retinoblastomas and can occur by different chromosomal mechanisms: (Fig. 4): mitotic nondisjunction with loss of the wild-type chromosome or duplication of the mutant chromosome, mitotic recombination between the retinoblastoma locus and the centromere, or gene conversion and deletion [14]. To establish the relative frequencies of those different events, Hagstrom et al. studied 158 cases of matched RNB and leukocyte DNA samples using polymorphic markers and found 64% of LOH [67]. Among those LOH cases, 7% were hemizygous after a deletion or nondisjunction without duplication and 93% homozygous (equal to about 55% chromosomal nondisjunction with duplication and 45% mitotic recombination). The authors also found no difference in the occurrence

Figure 4 Mechanisms leading to loss of the second RB1 allele [67]. The loss of the second RB1 allele can occur either after a small mutation or more frequently by the loss of the homologous normal allele (loss of heterozygosity, LOH). This event can occur by different chromosomal mechanisms (mitotic nondisjunction, deletion, and mitotic recombination). Hagstrom et al. have studied the frequencies of each event and their results are reported in this figure [67].

of LOH whether the patient was a man or a woman, whether the initial mutation was located in the germline or in a somatic cell, or whether the initial mutation arose from the paternal or maternal allele.

E.Low-Penetrance Retinoblastoma

About 10% of RNB families exhibit reduced penetrance (with unaffected obligate carriers) and variable expressivity of the disease (occurrence of RNB at an older age than usually expected in familial cases, with a high number of unilateral cases or benign retinocytomas) [68]. Some of those cases may be classified by mistake as unilateral sporadic cases. Characterization of those families and the underlying molecular mechanisms is relevant for accurate genetic counseling but also to further our understanding of the physiopathology of retinoblastoma and other cancers. To identify quantitatively low-penetrance retinoblastoma families, Lohmann et al. propose to calculate a disease : eye ratio (DER) that takes into account both penetrance and expressivity [69]. The DER is the number of eyes affected by RNB divided by the number of mutation carriers in the family. It is expected that families with complete penetrance and expressivity will typically have a ratio of 1.5 or higher, whereas most low-penetrance RNB families will have a DER inferior to 1.5.

It has been suggested that ‘‘weak’’ RB1 alleles could account for the low penetrance of retinoblastoma: the presence of those two alleles would be sufficient for tumor suppression, whereas the presence of only one (after loss of heterozygosity, leading to nondisjunction without reduplication or small intragenic mutation) would lead to retinoblastoma.