Clinical advances

Since the fourth edition of this publication, in 2006, the most significant addition to the clinical management of children with retinoblastoma, is an adaptation^2–4 of the Japanese experience in the use of intra-arterial (regional) chemotherapy.^5,6 Like many other new treatment approaches, the initial excitement has been tempered significantly as significant complications (including central retinal artery occlusion, local orbital recurrence and metastatic disease) have been observed. The pros and cons of this new approach to the treatment of retinoblastoma are discussed later in this chapter. Although RB1 gene mutation testing was recommended for everyone with a diagnosis of retinoblastoma regardless of age or circumstances of presentation (unilateral, bilateral, sporadic, or familial) Woo and Harbour reviewed 676 published second primary tumors (SPTs)⁷ the leading cause of death in heritable retinoblastoma patients.^8,9

Basic science advances

In the previous edition, we cited as advances: (1) the definition of the three-dimensional structure of the retinoblastoma protein (pRB) providing new insights into how it works; (2) the fact that pRb can simultaneously bind to multiple proteins, acting as a coupling factor to recruit proteins to specific genes; and (3) an improved understanding of low-penetrance retinoblastoma.¹⁰

Clinical genetics

Study of retinoblastoma has been critical to understanding the genetic basis of cancer in general. Although retinoblastoma is rare, occurring in approximately one in every 15 000 live births, the pathogenetic insights gleaned from retinoblastoma have profoundly affected our understanding of common cancers arising from lung, breast, prostate, and virtually every other site. This unique place of retinoblastoma in oncology derives from its very distinct pattern of inheritance and from the fact that the RB1 gene was not only the first human cancer gene to be cloned but was the first of a whole class of human cancer “suppressor” genes.

About 60% of all patients with retinoblastoma have a nonheritable form of the disease with a normal life expectancy if they are cured of the eye cancer. The average age at diagnosis is about 24 months, the eye tumor is unilateral, and the risk of other cancers is virtually indistinguishable from the normal population.¹¹ In contrast, the other 40% of patients, those with the RB1 germ line mutation, have a heritable cancer predisposition syndrome. The inheritance of a single inactive allele of the RB1 gene confers the predisposition to cancer (a dominant trait) but a second inactivating mutation must occur in at least one retinoblast for retinoblastoma to appear. Because the tumor requires inactivation of both copies of the RB1 gene, it is a recessive trait. However, in pedigrees, the tumor appears to be dominant because so many retinoblasts are at risk that the probability that at least one will get the required mutation to develop a tumor is at least 90–95%. A person with the cancer predisposing syndrome phenotype (RB1^+/−) will develop retinoblastoma with a 90–95% probability. Only 7–10% of retinoblastoma patients have a positive family history (someone else in the family with retinoblastoma). Hence, about 30% of patients with retinoblastoma have a new germinal mutation. The average age at diagnosis is younger than the nonheritable form – ranging from newborn to 12 months – and they are predisposed to a variety of cancers throughout life.¹² About 85% of heritable retinoblastoma patients develop multiple, bilateral eye tumors. Within the age range of 2–5 years, about 2–3% of these patients will develop a midline intracranial tumor usually involving the pineal or suprasellar region that histologically resembles retinoblastoma and has variably been referred to as a pinealoma, primitive neuroectodermal tumor (PNET), and trilateral retinoblastoma. It is important to recognize that the intracranial tumor is a primary cancer and not a metastasis from the eye tumor. The intracranial tumor is usually diagnosed before, together with, or within 2 years of the retinoblastoma, but it can be detected over 10 years later. Hereditary retinoblastoma patients have an elevated risk of osteosarcomas, soft tissue sarcomas and other mesenchymal tumors through their teenage years, melanomas and brain tumors through middle age, and epithelial malignancies such as lung and bladder cancer into later life. This elevated risk is much greater in the field of radiation treatment.

Genetic terminology

The genetic terminology used in the retinoblastoma field can be confusing. Clinicians erroneously may use the term unilateral to refer to the nonheritable form of retinoblastoma. However, this nomenclature is inappropriate, since about 15% of heritable cases develop tumors in one eye only.¹³ Consequently, epidemiologic studies that report tumor laterality but no information about the RB1 gene status probably are “contaminated” with unilateral, heritable patients and may give the erroneous impression that nonheritable patients have an elevated second cancer risk.⁸ The term sporadic is also commonly misused. Sporadic refers to a lack of family history but is not equivalent to nonheritable disease. About 90–93% of all retinoblastomas (heritable and nonheritable) are sporadic with no family history, meaning that they are present because of new mutations. In contradistinction, virtually all cases in a single family are hereditary, with the very rare exception of fortuitous and unfortunate familial aggregation of nonhereditary patients in the same family. Hence, nonheritable patients contain somatic or nongermline RB1 gene mutations (i.e., present only in somatic cells of the retina) are somatic mosaics. Hereditary patients carry germ line RB1 mutations (i.e., present in virtually all cells in the body, both somatic and germ line).

Molecular genetics of retinoblastoma

Because of the autosomal dominant inheritance pattern for retinoblastoma, the RB1 gene was assumed for many years to act in a dominant fashion.¹⁴ A major paradigm shift in the genetic understanding of retinoblastoma, and cancer in general, began with an enigmatic paper published in 1971 by Alfred Knudson, who proposed that retinoblastoma was caused by two mutational events: “In the dominantly inherited form of the disease, one mutation is inherited via the germ line and the second occurs in somatic cells. In the nonhereditary form, both mutations occur in somatic cells.”¹³ The major implication of this “two-hit theory” was that the RB1 gene functions in a recessive manner at the cellular level – an unprecedented suggestion at the time. Today, it is known that many cancer-causing genes are recessive or tumor suppressor genes.

The Knudson hypothesis languished for another decade due to a lack of scientific methods for identifying the RB1 gene. An early clue to the location of the RB1 gene was the recognition in the 1960s that a portion of a group D chromosome (13, 14 and 15) was occasionally deleted in retinoblastoma. Shortly after the Knudson paper, new chromosome banding techniques allowed chromosome 13 to be identified as the target of deletions.¹⁵ The smallest common deleted region was later mapped to chromosome 13q14.1 to q14.3.¹⁶ An enzyme with a measurable activity, esterase D, had been mapped to chromosome 13, and proved to be critical for linkage analysis in the era before recombinant DNA technology was readily available. Using classic deletion mapping, Sparkes and coworkers studied five patients with retinoblastoma and found that, in all five, esterase D activity was only 50% of normal. These data suggested that the loci for retinoblastoma and esterase D genes were both within the deleted segment of chromosome 13q.¹⁷ Based on the fact that the chromosome 13 deletion and the esterase D locus were tightly linked to retinoblastoma in multiple families with clinically and pathologically indistinguishable disease, Murphree and Benedict argued that there was probably a single RB1 locus.¹⁸ Further, retinoblastoma tissue from a nonhereditary unilateral patient was also found to contain a 13q14 deletion, suggesting that all forms of retinoblastoma involve the same gene on chromosome 13q14.¹⁹ The current consensus is that there is indeed a single retinoblastoma locus, RB1, that is mutated in all forms of retinoblastoma.

Meanwhile, a body of scientific work was accumulating to support the notion that RB1 is a recessive gene that is lost during tumorigenesis. Benedict and coworkers examined a familial retinoblastoma patient and found a 50% reduction of esterase D in normal cells and a complete absence of the enzyme activity in retinoblastoma cells.²⁰ Dryja and colleagues used DNA fragments from 13q14 to show that homozygous deletions could occasionally be identified in retinoblastoma tissue.²¹ In two landmark papers, Cavenee et al. proved beyond doubt the recessive nature of the RB1 gene and popularized the use of loss of heterozygosity analysis. By comparing DNA from normal and tumor tissue, they found that the region around the RB1 gene was frequently “reduced to homozygosity” in retinoblastoma tissue. In other words, one copy of the region around RB was lost during tumorigenesis. In 1985, Cavenee et al. went a step further by showing that, in heritable cases, the germ line copy of 13q (carrying the mutant RB1) that was passed among affected family members was always the one that was retained in the tumor.²² The chromosomal mechanisms involved in reduction to homozygosity on 13q14 are shown in Figure 128.1.

Fig. 128.1 Chromosomal mechanisms of loss of heterozygosity (or reduction to homozygosity) that cause the loss of the second RB gene allele in retinoblastoma.

Another breakthrough in the hunt for RB1 was the cloning of the esterase D gene.^23,24 In the early days of recombinant DNA technology, cloning a gene usually required an activity that could be assayed, and esterase D provided such an activity. Once the esterase D gene was cloned, its DNA sequence was harnessed and could be used to probe adjacent stretches of chromosomal DNA to identify nearby genes. An intensive search ensued for the RB1 gene, and as luck would have it, the esterase D gene proved to be located very nearby the RB1 gene. A few short months following the cloning of esterase D, the search culminated in the discovery of a large gene that contained deletions in many retinoblastomas.^25,26 In addition, the mRNA transcript from this gene was either missing or abnormal in size in most retinoblastomas. Even though some early workers questioned whether this gene was indeed RB1,²⁷ further work has confirmed that this is the gene that is mutated in retinoblastoma. For example, re-introduction of RB1 gene into retinoblastoma cells and other RB1-deficient tumors²⁸ suppressed the neoplastic phenotype, indicating that the RB1 gene was indeed a tumor suppressor. Thus, by the early 1990s, there was a basic molecular under standing of how retinoblastoma is inherited. Transmission of an inactive copy of RB1 imparts the predisposition to retinoblastoma. Inactivation of the second copy because of the random background mutation rate leads to tumor development.

The RB1 gene

The RB1 gene (Fig. 128.2) contains 27 exons spanning over 200 kilobases of DNA. The 5′ end of the gene is oriented toward the centromere of chromosome 13. The promoter region lacks a typical TATA box but contains a CpG-rich rich region, or CpG island. Germ line mutations in the RB1 gene are distributed throughout the gene with mutational hotspots at CpG dinucleotides.^29,30 Less than 10% of retinoblastoma patients have a constitutional chromosome 13q abnormality (usually a deletion) that can be detected by karyotyping.^31,32 Deletions that are more extensive can be associated with the 13q-syndrome, with features such as growth and mental retardation, facial dysmorphism, microcephaly, skeletal anomalies, and genitourinary abnormalities. About 15–20% of germinal mutations are too small to detect by cytogenetics but can be detected with techniques for analyzing gross DNA rearrangements (e.g., Southern blot). The remainder of RB1 gene mutations are small alterations involving one or a few nucleotides that can only be detected by high resolution methods and direct sequencing.^33,34 The majority of these mutations are frameshift and nonsense mutations, inframe and missense mutations, splicing mutations, and mutations in noncoding regions that result in a truncated, unstable protein product (Table 128.1).³⁵ The possibility of using truncated pRb protein mutants as the basis for genetic testing to detect germ line RB1 mutations was proposed several years ago,²⁹ and recently shown to be feasible.³⁶ Interestingly, most new germ line RB1 mutations are of paternal origin, suggesting that the gene is more susceptible to mutation during spermatogenesis rather than oogenesis.³⁷

Fig. 128.2 RB gene, mRNA transcript, and protein.

Mutation of the second RB1 allele is typically due to chromosomal aberrations, usually recognized as loss of heterozygosity (LOH) by polymorphism analysis.²² These mutations occur at a much higher rate than the first, germ line mutation (10⁻³ as compared with 10⁻⁷ for the first mutation).³⁸ The most common mechanism leading to LOH is mitotic recombination (50%),³⁹ followed by nondisjunction with or without subsequent reduplication (~40%). Other mechanisms include small deletions and gene conversions. Therapeutic exposure to ionizing radiation induces DNA damage and increases the risk of LOH, and hence tumorigenesis, within the field of radiation.

Advances in rapid genetic testing have allowed for preimplantation genetic diagnosis.⁴⁰ This procedure allows a couple in which one parent has a known germline mutation to undergo in vitro fertilization where the prior to implantation, the embryo’s are tested for the mutation using single cell polymerase chain reaction and rapid DNA sequencing of the exon in question. Only those embryos which are mutation free are implanted back into the mother. This has been successfully performed although due to the expertise and technology required, it is not a widely available technique.

Low penetrance retinoblastoma

The term penetrance refers to the frequency that a heritable disease is manifest in offspring of affected individuals. The word expressivity refers to the variability of clinical manifestations in affected individuals. For example, patients with heritable retinoblastoma who develop only unilateral eye disease manifest reduced expressivity. In general, reduced penetrance and expressivity tend to segregate in the same families.⁴¹ Overall, the penetrance of retinoblastoma is about 80–90%, but this represents a heterogeneous group of high-penetrance and low-penetrance families.⁴¹ The diseased-eye ratio (the ratio of the number of eyes containing tumors to the number of mutation carriers in a family) was devised to quantitatively identify low penetrance retinoblastoma families by taking into account both penetrance and expressivity.³⁵ Most low penetrance families have diseased-eye ratios <1.5, whereas families with complete penetrance typically have diseased-eye ratios of ≥1.5. Initially, investigators postulated that low penetrance retinoblastoma may be due to immunologic factors, DNA methylation, epigenetic mechanisms, delayed mutation, host resistance factors, a second retinoblastoma locus, or modulator genes.^41–44 However, recent research has shown that most low-penetrance retinoblastoma results from mutations at the RB1 gene locus that result in an pRb protein with reduced activity.^45–47 One of the most common low penetrance mutations is a missense alteration at codon 661 (exon 20).^45,48 Other reported low penetrance mutations include 3-bp deletion in exon 16 that results in the deletion of Asn480,50 a 4-kb deletion involving exons 24 and 25,54 and a splicing mutation at the last base of exon 21.⁴⁷ Insights gained from these low penetrant RB1 mutants have added much to our understanding of how the retinoblastoma protein works.¹⁰

RB1 gene mutations in other tumors

Not surprisingly, RB1 gene mutations occur frequently in tumors linked to retinoblastoma, such as osteosarcomas, soft tissue sarcomas and other mesenchymal tumors.^49–51 However, investigators were surprised to find that the RB1 gene is also mutated frequently in some common adult malignancies such as breast, lung, and prostate cancer.^52–54 As our understanding of the RB1 gene and its protein product have progressed, it has become clear that RB1 is disrupted in most human cancers, either by mutation of the gene or, more commonly, by functional inactivation of the protein.⁵⁵

The role of the retinoblastoma protein in tumor suppression

The retinoblastoma protein

The RB1 gene encodes a 4.7 kilobase messenger RNA transcript that produces a protein of 110 kD (kilodaltons) and 928 amino acids (Fig. 128.2). The pRb protein is phosphorylated in a cell cycle-dependent manner and localizes to the nucleus.^56–58 The hypophosphorylated form predominates in quiescent and differentiating cells, whereas the hyperphosphorylated species accumulates in cycling cells as they enter DNA synthesis (S) phase (Fig. 128.3).^59–63 The hypophosphorylated form of pRb binds several viral oncoproteins, including SV40 large T antigen,⁶⁴ adenoviral E1a,⁶⁵ and human papillomavirus E7.⁶⁶ When bound to pRb, these oncoproteins stimulate cell division. Taken together, these findings provide evidence that hypophosphorylated pRb is important in negatively regulating the cell cycle, and that this inhibitory activity can be thwarted either by phosphorylation or viral oncoprotein binding. Further work has shown that the major cell cycle function of pRb is to inhibit the transition of cells out of gap 1 (G1) phase into S phase.⁶⁷ However, pRb may also have roles in other cell cycle phases.^68,69

Fig. 128.3 Role of the Rb protein in the cell cycle and apoptosis.

A major breakthrough in understanding how pRb regulates the cell cycle was the observation that pRb binds to members of the E2F transcription factor family (referred to here as E2F).^70–72 Further work has shown that pRb function is largely dependent on interactions with E2F.⁷³ E2F sites are found in the promoters of many genes that are important for cell cycle progression, and pRb represses transcription of these genes through its interaction with E2F.^74–77 Since E2F (but not pRb) has a DNA binding domain, the pRb-E2F association would explain how pRb is brought to specific DNA elements to exert its effect. Most E2Fs have a transactivation domain that stimulates expression of genes containing E2F binding sites in their promoters. pRb binds E2F within the transactivation domain,^78,79 thereby masking its activity. Since E2F activates genes involved in cell division,^74,80 inhibition of E2F provided a mechanistic explanation for how pRb inhibited cell division. However, the picture became more complicated with the recognition that pRb has intrinsic or “active” transcriptional repressor activity and is able to block the expression of genes when artificially brought to promoters by proteins other than E2F.⁸¹ These findings suggest a complex relationship between pRb and E2F. In some situations, pRb inhibits genes by simply masking the E2F transactivation domain, whereas in other situations E2F serves as a “courier” to deliver pRb to specific genes for active repression.

The importance of active repression by pRb was pointed out in several studies that showed that this activity is required for pRb to inhibit the G1-to-S phase transition of the cell cycle.^82,83 But how does pRb actively represses transcription? In a series of landmark papers, several groups showed that pRb binds to and recruits to promoters proteins that alter chromatin structure, such as histone deacetylases,^83–85 SWI-SNF ATPases,^86–88 DNA methyltransferases,⁸⁹ polycomb complexes,⁹⁰ and histone methylases.⁹¹ Alteration of local chromatin structure into a restricted conformation prevents access by the transcriptional machinery, thereby inhibiting expression, whereas dynamic reorganization of chromatin into an open configuration allows gene transcription. Depending on the nature of the chromatin-remodeling complex that pRb recruits, the cell cycle inhibition can be temporary, as occurs during the quiescent period between cell divisions, or permanent, as occurs during cell differentiation and senescence.⁹²

Studies of the tertiary (three-dimensional) structure of the pRb protein have provided insights into how the protein performs these complex functions (Fig. 128.2). The central region of the pRb protein contains the A box and B box, which are highly conserved from human to plants. These regions interact with each other along an extended interdomain interface to form the A–B pocket. The pocket is critical for the tumor suppressor function of pRb, and is disrupted by most germ line mutations in hereditary retinoblastoma patients and somatic mutations in tumors.^29,93 The pocket is required for binding to E2F, chromatin remodeling enzymes, viral oncoproteins, and other molecules. Many pRb-binding proteins contain an LxCxE (leucine – variable amino acid – cysteine – variable amino acid – glutamic acid) binding motif, which has been shown by crystallographic studies to be located in the B box.⁴ The B box does not assume an active confirmation unless bound to the A box, thereby providing an explanation for why both boxes of the pocket domain are required for pRb activity. Interestingly, E2F does not contain the LxCxE motif and binds pRb at a distinct site (Fig. 128.3). Recent crystallographic studies confirmed that the E2F binding site, located at the interface between the A and B boxes, is distinct from the LxCxE site.⁹⁴ These findings provide a structural explanation for how pRb simultaneously binds E2F and chromatin remodeling proteins, many of which interact with pRb through the LxCxE site.⁸⁸ The current molecular view is that pRb orchestrates the assembly of multi-protein complexes, which are then recruited to specific promoters by E2F, where they control access of the transcriptional machinery. Thus, pRb regulates in a dynamic and integrated manner the expression of specific genes involved in cell division, differentiation and apoptosis.⁹⁵

The carboxy-terminal end of the pRb protein is also critical for its function. This region is required, along with the pocket, for binding to E2F.⁹⁶ In addition, most of the phosphorylation sites that seem to be critical for regulating pRb activity are located in the carboxy-terminus.⁹⁷ In fact, recent work has shown that phosphorylation of pRb on the C-terminus initiates a novel intramolecular interaction that progressively strips pRb of activity as the cell moves through the cell cycle.⁹⁸ The carboxy-terminal region also contains binding sites for the oncoproteins c-abl and MDM2.^99,100 The tyrosine kinase activity of c-abl is blocked when it is complexed with pRb,⁹⁹ and this interaction appears to be important for Rb-mediated growth suppression.¹⁰¹ Besides directly blocking c-abl, the C-terminal region also appears to participate in the assembly of multimeric complexes containing pRb, E2F, c-abl and potentially other proteins.¹⁰² The importance of the pRb-MDM2 interaction is less clear. MDM2 interacts with the p53 tumor suppressor protein and opposes its proapoptotic activity by repressing p53 transcriptional activation and by mediating its degradation.^103,104 While initial results demonstrated that MDM2 blocks pRb function, more recent studies have shown that pRb can form a trimeric complex with MDM2 and p53 and thereby block the antiapoptotic activity of MDM2 by preventing the degradation of p53.¹⁰⁵

The function of the amino-terminal end of the pRb protein remains less clear. This region contains phosphorylation sites that may regulate pRb activity. Additionally, this region interacts with several proteins, including a replication-licensing factor (MCM7),¹⁰⁶ a novel G2/M cycle-regulated kinase,¹⁰⁷ and other proteins.¹⁰⁸ However, the function of these interactions has not been established. RB1 ± mice that develop pituitary tumors due to loss of Rb failed to be “rescued” by expression of a mutant form of pRb that lacked the amino-terminus, although the onset of tumors was delayed.¹⁰⁹ In other experiments, tumor suppression by pRb was actually enhanced when the amino-terminal region was removed.^110–112 Thus, the amino-terminus appears to contribute only weakly to the overall tumor suppressor activity of pRb.

Not only was the RB gene the first identified tumor suppressor gene, but the Rb pathway was the first, and still one of the most important, tumor suppressor pathways to be elucidated in human cancer.⁵⁵ As described above, the pRb protein is critical for regulation of the cell cycle, as well as senescence, differentiation and apoptosis, all of which are deregulated during cancer formation.^{67,92,113–120} While the pRb pathway is deregulated in virtually all cancers,⁵⁵ the RB1 gene is mutated in only a limited spectrum of cancers. In these other cancers, the pRb protein is inactivated by maintaining it in a hyperphosphorylated state. This occurs by deregulating the pRb pathway, which controls the phosphorylation state of pRb through kinases and kinase inhibitors.

The RB-E2F regulatory network

The complexity of the pRb tumor suppressor pathway is greatly increased by the fact that there are two other members of the pRb pocket protein family – p107 and p130 – as well as seven or more members of the E2F family.⁹⁵ The genes for p107 (RBL1) and p130 (RBL2) have been cloned and mapped to human chromosomal regions 20q11.2 and 16q12.2, respectively.^131,132 The p107 and p130 proteins share extensive homology with pRb within the pocket domain and are involved in cell cycle regulation.¹³³ While it is still not clear why so many pRb and E2F family members are necessary, recent work suggests that specific pRb family members interact preferentially with specific E2F family members during precise phases of the cell cycle. For example, pRb interacts primarily with E2F1–3 and is most active at the G1-to-S phase transition. In contrast, p130 interacts primarily with E2F4 and E2F5 and is most active in G0 – the quiescent phase of the cell cycle.⁹⁵ A mechanistic picture is emerging in which the pRb-E2F network regulates gene expression through complex and context-dependent interactions between pocket proteins, E2F family members, and cofactors such as chromatin remodeling.

Molecular pathogenesis of retinoblastoma

There is now overwhelming evidence that supports the hypothesis that mutational inactivation of the RB1 gene is the initiating event in retinoblastoma. However, there remain many unanswered questions about the molecular pathogenesis of this tumor. Since pRb is important for regulating normal cell growth and differentiation throughout the body, why does germ line mutation of the RB1 gene predispose primarily to the rare eye tumor? This question was once quite enigmatic but is now becoming clearer. Indeed, carriers of RB1 mutations are predisposed to a whole range of tumors, but the age of susceptibility is different for each tumor type. Whereas retinoblastomas occur mostly from birth to 5 years of age, mesenchymal tumors tend to occur in the teenage years, and melanomas peak in a slightly older age group. A recent report clearly shows that common epithelial tumors also occur at increased frequency in retinoblastoma survivors, but this effect is only seen in individuals beyond 40 years old.¹³⁴ Thus, it seems likely that loss of RB1 predisposes to a broad range of neoplasms, each requiring different numbers and types of “hits” for tumorigenesis to become manifest; the fewer “hits” required, the earlier in life the tumor begins to appear in patients.

While mutation of the RB1 gene is clearly necessary for retinoblastoma formation, is inactivation of RB alone sufficient for tumorigenesis? Some investigators have argued additional mutations in other genes must be required based on the following observations. First, deletion of the RB1 gene in normal cells leads to apoptosis, rather than tumor formation, because loss of pRb triggers a p53-mediated apoptotic response.¹³⁵ This presumably explains why most cancers contain mutations in both the pRb and p53 pathways.^92,136 Second, retinoblastomas cannot be produced in mice unless both pRb and p53 are inactivated.¹³⁷ This led to a search for mutations in p53 or other pro-apoptotic genes. Interestingly, however, p53 is rarely mutated in human retinoblastoma,¹³⁸ and no other apoptotic genes have been convincingly linked to retinoblastoma. Finally, cytogenetic alterations on other chromosomes (e.g., 6p) are frequently observed in retinoblastomas,¹³⁹ potentially suggesting the presence of other retinoblastoma-associated genes located at those chromosomal regions.¹⁴⁰

Recent work has shown that MDM2 and MDMx may play an important role in attenuating the normal apoptotic events that should occur in an RB1^−/− state. MDM2 targets p53 for ubiquitin mediated proteolysis and is also a downstream target of p53. This creates an auto-feedback loop that maintains p53 at a low level. Small molecule inhibition of MDM2 has been shown to lead increased levels of p53 in retinoblastoma cell lines leading to p53 mediated apoptosis.^141,142

While the existence of a “second RB gene” cannot be confirmed or refuted at this time, the strong autosomal dominant inheritance pattern of retinoblastoma indicates that the RB1 gene mutations are clearly rate-limiting, so any other necessary events must occur at such a high spontaneous rate that they do not affect the clinical inheritance pattern. The emphasis on mouse models to understand human retinoblastoma may have limitations. Naturally occurring retinoblastomas are virtually nonexistent in mice and other animals, suggesting that the developing human retina may contain molecular, cellular and anatomic features that make it uniquely susceptible to this tumor.

An alternative explanation for retinoblastoma pathogenesis is simply that retinal progenitor cells pass through a window of susceptibility prior to terminal differentiation in which loss of pRb leads to a differentiation defect, rather than an apoptotic response, which leads to an accumulation of proliferating embryonic retinal cells. This hypothesis does not require the conjectural existence of additional “retinoblastoma genes” and satisfactorily accounts for developmental and clinical observations. pRb is indeed required in a cell-autonomous manner for appropriate cell cycle exit and differentiation of retinal progenitor cells.¹⁴³ Further, the topographic distribution of retinal tumors parallels the pattern of retinal differentiation. The retina differentiates in a posterior-to-anterior wave from the posterior pole to the ora serrata.¹⁴⁴ Interestingly, the chronological development of retinoblastomas follows this same pattern, with earlier tumors occurring posteriorly and later tumors occurring more peripherally.¹⁴⁵ Retinal progenitor cells that retain the capacity to differentiate into photoreceptors, neurons, and glia can be identified in the retina until after birth,¹⁴⁶ suggesting a window of susceptibility from fetal week 12 until 4–5 years of age for the bi-allelic loss of the RB1 gene to have the possibility of producing retinoblastoma.

Another controversy regards the cell of origin of retinoblastoma. Retinoblastomas derive from cells of the immature neuroepithelial inner layer of the optic cup that have the potential to differentiate into rod and cone photoreceptors and Müller cells.¹⁴⁷ Several studies have documented the presence of cone-specific markers, such as transducin, cone photopigments (red and green opsin), and cone phosphodiesterase in retinoblastomas.¹⁴⁸ However, since cone differentiation may represent a “default” pathway in the absence of normal signaling,¹⁴⁹ it is unclear whether retinoblastomas arise from neuroblasts that are already committed to the cone lineage, or whether retinoblasts that lose pRb are unable to differentiate along their appropriate lineage and are subsequently directed into the cone pathway. Recent work has identified an additional association between cone cell precursors and mature retinoblastoma tumors. Cone cell precursors, but not mature cone cells, express pRb. In addition, cone cell precursors also express MDM2, which is a transcriptional target of Trβ2 that is also transiently expressed in cone cell precursors. Taken together, loss of pRb in a cone cell precursor that is already expressing MDM2 as part of its developmental program could plausibly lead to transformation and would be consistent with this being a potential cell of origin.

Terminology

Retinoblastoma is a complicated subject that can be challenging for both medical professionals and affected families to fully understand. Imprecise terminology compounds confusion. To avoid unnecessary complexity, it is important to use precise terminology when thinking or talking about retinoblastoma.

One of the most subtly confusing terms in the retinoblastoma lexicon is “bilateral retinoblastoma.” The term should be used only to describe a phenotype, i.e., a patient who has the disease in both eyes. When the intent is to define a genotype – the whole group or class of patients with a germ line mutation, a much clearer term is “heritable or hereditary” retinoblastoma. All patients with bilateral disease do carry the RB1 germ line mutation but not all people who carry the RB1 mutation have bilateral disease. The phenotype (clinical presentation of the disease), will depend upon whether the second tumorigenic RB1 mutation occurs, and cancer is initiated, in both eyes (bilateral), only one eye (unilateral) or neither eye (unaffected gene carrier).

The term “hereditary” means “having inherited status” and is correctly used to describe a patient or family in which the germ line mutation was passed down from a previous generation. A new sporadic bilateral case of retinoblastoma when neither parent has the RB1 mutation is more precisely referred to as “heritable” rather than “hereditary,” although many do not make this distinction. A unilaterally affected child who has one parent with a known RB1 mutation is clearly both heritable and hereditary but obviously not “bilateral.”

Overview of retinoblastoma

Retinoblastoma may be present at birth. In most cases, it is diagnosed between birth and 5 years of age. The tumor may involve one or both eyes. Each tumor arises independently. Early tumors <3 mm in greatest diameter may appear as discrete round, gray-white masses without intrinsic vascularity usually in the posterior pole of the eye. In dim light when the pupil dilates naturally, in the early days of the tumor natural history, parents may see light reflected by the tumor causing a “cat’s eye” reflection in the pupil or a yellow glow rather than the expected “red eye” on flash photographs (Fig. 128.4). Less commonly, more advanced intraocular retinoblastoma may manifest as the lack of a red pupillary reflex and may appear black compared with the red glow from the other pupil when viewed through a direct ophthalmoscope. This is caused by either blood in the vitreous from tumor necrosis or when there is a total retinal detachment and the retina is pushed up behind the lens. When the tumor destroys central vision early in the course of the disease, and binocularity is lost, the presenting sign may be strabismus.

Fig. 128.4 (A) Patient with leukocoria in right eye and a white mass in the pupil. (B) “Cat’s eye” reflection in the pupil seen easily in a flash photograph.

Initially tumors grow rapidly in the retina. As they expand, spontaneous mutations creating clones of cells each with a few more malignant characteristics including anchorage independence leading to vitreous and subretinal seeding is the simplest explanation that fits the observed findings. Eventually, the expanding mass can cause retinal detachment, neovascular glaucoma and/or intraocular hemorrhage before it spreads beyond the eye. Some tumors may spontaneous involute, activating a process that will result in phthisis. If left untreated, most intraocular retinoblastomas will expand directly into the optic nerve, the brain and/or orbit, spread via hematogenous dissemination to bone marrow, bone, and other organs. Enucleation was the recommended treatment in James Wardrop’s original clinical description of retinoblastoma in 1809 and remains a superb treatment option today, especially for advanced unilateral disease.

Epidemiology

Since the last edition of this book was published, striking differences between the relative prevalence of unilateral nonheritable and heritable disease has come to wide attention in part due to the important work of Ian McGrath and colleagues at the International Network of Cancer Treatment and Research (see the INCTR website at: http://www.inctr.org/ accessed December 9, 2011). Orjuela and her colleagues at Columbia University in New York and at the Instituto National de Pediatria in Mexico City have suggested deficiencies in maternal diet during pregnancy.¹⁵⁰ While the worldwide incidence of the heritable form of the disease is relatively constant, such is not the case for unilateral sporadic non-heritable retinoblastoma.

In the USA, retinoblastoma is the sixth most common solid childhood tumor. In developing tropical and subtropical regions of the world, including Central Africa, Southern Asia, and Central America retinoblastoma is at or near the top of the list as the single most common solid tumor of childhood.^151–168 The difference is due for the most part to an increased incidence of unilateral non-heritable disease.

The overall incidence (heritable and nonheritable cases combined) in developed countries is most commonly reported as the number of cases of retinoblastoma diagnosed in a selected period per live births for that period.¹⁶⁹ Although this index does not correctly reflect the population at risk, since children continue to be at risk until 5 years of age and beyond,¹⁷⁰ it does present some uniformity for comparison of the frequency of the disease across populations.

Earlier studies of the incidence of retinoblastoma in the USA report rates of less than 1 in 20 000. From 1970 on, the prevalence (number of cases per number of persons in the at-risk age group) is consistently reported at around 11 cases per million children under 5 years of age,¹⁷¹ or equivalent to an incidence of 1/18 000 live births.¹⁷² Because there is no complete retinoblastoma registry in North America, the exact incidence is unknown. However, the estimate retinoblastoma specialists usually agree upon is about 300 new cases per year in North America. If the estimate were based on the 2007 Census data for the number of annual births in the USA (4 242 000), the incidence would be 1 : 17 000. This number fits closely with studies reported from New Zealand¹⁶⁹ and Sweden,¹⁷³ both taken from national registries, as well as a regional study from Australia,¹⁷⁴ all show incidence rates of 1/17 000–1/18 000 live births.

The incidence of heritable retinoblastoma among the various populations of the world is remarkably constant, providing strong evidence that environmental influences play little role in the etiology of the hereditary form of this tumor.^169,175 Buckley provided evidence that environmental factors play little role in the hereditary form of cancer in very young children.¹⁷⁶ Bunin and colleagues from nine North American centers have demonstrated that medical radiation exposure (from lower GI series) prior to conception significantly increased the risk that a child develop sporadic retinoblastoma from a new germline mutation has an extensive epidemiological study underway evaluating the role, if any, of environmental exposure of the father in sporadic heritable retinoblastoma.³¹

In marked contrast to the uniform incidence of heritable retinoblastoma worldwide, as noted above, there are striking geographic differences in the incidence of the nonheritable, unilateral form of retinoblastoma from one region of the world to another. The differences in overall incidence are due entirely to the excess of nonheritable, unilateral cases of retinoblastoma. There exists a possibility that the increased incidence of nonheritable retinoblastoma in the poorer, tropical and subtropical regions of the world is due to a viral etiology (possibly human papilloma virus, HPV),¹⁷⁷ although convincing proof is still lacking. This group has also found that nonheritable retinoblastoma seems to be statistically more common when pregnant mothers eat a diet deficient in fruits and vegetables. A major prospective epidemiological study has been funded to allow this group to evaluate possible dietary influences during pregnancy on the increased incidence of nonheritable retinoblastoma in developing subtropical countries.

Advanced paternal age is unequivocally associated with new sporadic germ line mutations and sporadic heritable retinoblastoma.^178–184 Excess cancer in relatives¹⁸⁵ is also a common finding in solid childhood tumors, particularly retinoblastoma. Some 80–85% of new heritable tumors preferentially retained the paternal allele (i.e., the mutated allele) and lost the normal maternal allele as the result of a chromosomal error at mitosis.^180,186 These data suggest that new germ line RB1 mutations arise more frequently during spermatogenesis than during oogenesis.

Natural history of intraocular retinoblastoma

If the assumption that retinoblastoma tumors become ophthalmoscopically apparent in the retina within a short time after inactivation of the second wild-type allele at the RB1 locus, then the conventional thinking that retinoblastoma can have its initiating event as late as the 2nd or 3rd birthdays. However, if deficiency of maternal diet can affect the incidence of unilateral nonheritable retinoblastoma, then evidence supports the view that all retinoblastoma mutations including somatic mutations occur during fetal life. Other factors may be at play in the observed differences in the age at clinical diagnosis between heritable and nonheritable retinoblastoma.

Regardless of when the cellular events that lead to retinoblastoma take place, the earliest physical appearance of the tumor is fixed by the fact that all cells in the tiny tumor focus are identical, i.e., they are daughter cells of the original “founder” cell with the same genes and the same growth rate. A new tumor will expand symmetrically as a round or hemispherical lesion that is homogenous. In its earliest manifestation, retinoblastoma resembles bacterial colonies on an agar plate (Fig. 128.5). Because tumor growth begins with a single immortalized retinoblast, all intraocular retinoblastomas are initially confined to the retina.

Fig. 128.5 Solitary tumor that is symmetrically round and homogeneous due to a clone of “founder” cells with identical genes and growth rate.

More than 25 years ago, aliquots of aqueous from eyes containing intraocular tumors (primary or metastatic) were shown to induce angiogenesis on the chick allantoic membrane even if the aliquot was taken before tumor was clinically apparent.¹⁸⁷ Tumor angiogenesis is essential if tumors are to expand. Conversely, avascularity severely restricts the potential growth of tumors.¹⁸⁸ Intraretinal lesions of retinoblastoma reach that growth restriction at a diameter of between 1 and 2 mm if they remain dependent on diffusion of nutrients and oxygen from the choroid. Vessels grow into the tumor only in response to growth factors generated from upregulated growth factor genes and downregulated angiogenesis inhibitory genes. A physical marker for the “turned-on” angiogenesis in any particular tumor may rest in our ability to accurately estimate the number of new vessels existing in a tumor. This relative tumor vascular density (RTVD) can be demonstrated by CD-34 staining for vascular endothelium in special stained histopathology sections. The RTVD is statistically greater in ocular tumors of patients with metastatic disease.^189,190 In the February 2004 report from Marback and colleagues, tumor vessel density was also greater in tumors that have invaded the choroid and optic nerve, previously identified histological risk factors for metastatic disease.¹⁸⁹

Normal cells adhere to each other and the extracellular matrix through a series of specialized molecules. An important component of the “progression of malignancy” process of any tumor cell is loss of this cellular adhesion or anchorage dependence. A tumor suppressor gene, PTEN, is mutated in many advanced tumors.¹⁹¹ When a normal wild-type copy of the gene is introduced into tumor cells the exogenous PTEN inhibits tumor cells’ ability to grow anchorage independently. The product of this tumor suppressor gene was able to revert one of the typical properties of tumor cells to normal. The controls that would ordinarily shunt a cell to apoptosis or the “programmed cell death pathway” if it lost its attachment to the extracellular matrix may be eliminated in the “progression of malignancy” process.