Ординатура / Офтальмология / Английские материалы / Ocular Oncology_Albert, Polans_2003

testing the blood DNA for each of those mutations. In this way, a negative result generally rules out a heritable retinoblastoma in the proband, which is useful for risk prediction for the relatives. However, a small possibility of mosaicism remains, as discussed below.

A.Spectrum of RB1 Mutant Alleles

1.Expeditious Strategies for Screening for RB1 Mutations

RB1 mutations occur throughout the gene and the majority are ‘‘null,’’ resulting in no detectable pRB. More than 50% of RB1 mutations have been found once only, and the majority of germline mutations have originated with the proband. Therefore the job of identifying the RB1 mutation in each family is onerous. To detect the mutation in each family, a series of assays and different technologies are required. Various scanning techniques help identify suspect exons, and follow-up sequencing detects approximately 60% of the mutations. Somatic mutations in tumors that involve methylation of the promoter occur in about 10% of tumors [45]. To achieve 90% sensitivity of identifying the RB1 mutation in each family, we have added quantitative multiplex polymerase chain reaction (QM-PCR) to detect copy number changes of single and multiple exons. The types of mutations include single base substitutions, resulting in amino acid substitutions or altered splice sites, small deletions and large ones up to deletion of the whole gene, exonic deletions, intraexon deletions, and insertions and methylation of promoter regions. RB1 mutations are spread throughout the gene including the promoter, occurring in most exons and some introns, with no significant hot spots for null mutations.

2.Our Results for RB1 Mutation Screening

We have identified more than 450 mutant RB1 alleles [45a]. In bilateral probands, 93% of the mutations have been null, 6% are in-frame, and 1% are promoter. The RB1 germline mutations in unilateral probands (both familial and sporadic) are 57% null only, while 40% are in-frame and 3% are promoter. The reduced expressivity manifests as fewer than expected tumors, therefore more often occurring as unilateral rather than bilateral disease. This correlates with the presence of ‘‘weak’’ RB1 mutant alleles that retain enough activity to prevent some but not all retinoblastoma tumors. The mutant alleles that have been identified in retinoblastoma tumors but not in blood, therefore presumably somatic, are 88% null, 3% inframe, and 8% promoter methylation. In summary, the overall rate to successfully complete the clinical analysis (persons diagnosed/persons analyzed) was 86% for families with unilateral germline mutations, 89% for families with bilateral probands, and 82% for unilateral retinoblastoma probands with no family history for which both mutant alleles must be found in each of the tumors.

For sporadic retinoblastoma tumors, the second allele was mutated by loss of heterozygosity (LOH) in 52%. The likelihood that the M2 event would be LOH varied with the type of M1 event. Only 22 (27%) of the tumors with M1 whole gene and exonic deletions, respectively, showed LOH, while 89% of tumors with methylation of the promoter showed LOH.

Some ‘‘recurrent’’ mutations, including large deletions and methylation of the promoter, have not been defined at a nucleotide level. Thirteen true recurrent point mutations occurred four or more times in our dataset, all resulting from CT transitions at CpG dinucleotides because of deamination of 5-methylcytosine [46,47]. Twelve affect arginine codons, and one creates a splice mutation. There is a true ‘‘hot spot’’ for missense mutations, involving the A/B ‘‘pocket’’ domain of pRB, which is critical in the interaction of pRB with the transcription factor E2F [48]. Presumably, missense mutations in the remainder of RB1 are insufficient to initiate retinoblastoma.

B.Practical Clinical Detection

1.Clinical Importance of RB1 Mutation Identification

Precise identification of the RB1 mutations that account for retinoblastoma in each family enhances the quality of clinical management of the affected patient and relatives at risk. Individuals with RB1 mutations have a higher-than-normal lifetime risk to develop additional cancers [49]. Exposure to external beam radiation further increases that risk. In the absence of knowledge of the precise RB1 mutation in the family, children at risk to have retinoblastoma undergo a series of clinical examinations, including examination under anesthetic (EUA), in order to diagnose and treat tumors as early as possible. If the RB1 gene status of the proband has been determined by molecular testing, only those relatives with the mutation require clinical surveillance, while those proven not to be carriers require no further testing. The direct costs of molecular testing are significantly less than conventional clinical examinations for each family [50]. If a family’s mutation is known, prenatal molecular testing allows careful planning of perinatal management for infants with RB1 mutations. We have delivered five infants prematurely to facilitate early treatment of macular tumors. Two of the babies had macular tumors at 35 weeks gestation, and all of the children had bilateral retinoblastoma within the first 3 months of age.

2.‘‘Best Practice’’ Guidelines for RB1 Mutation Identification

Despite clear clinical benefits of molecular testing for retinoblastoma, mutation detection in the RB1 gene has not been widely implemented. The wide variety of inactivating mutations, and their distribution along the entire length of the gene without any significant hot spots, have made RB1 mutation identification difficult and expensive. In addition, the heterogeneity of the RB1 mutation spectrum suggests that no one single mutation identification technology could be totally sensitive and fully efficient. Several testing laboratories have focused on clinical RB1 mutation testing and refined their technologies and interpretation to enable implementation in practice. These laboratories include the Institut fu¨r Humangenetik, Essen, Germany (D. Lohmann), and Retinoblastoma Solutions (http://www.solutionsbysequence. com). The European Molecular Quality Network has already formalized the validation of RB1 testing and published this on the EMQN website (http:// www.emqn.org/guidelines/rb.php).

C.Origin of RB1 Mutations and Mosaicism

1.Frequent Mutation of the Paternal RB1 Allele

The majority of children with heritable retinoblastoma have newly acquired RB1 mutations. Most frequently, it is the paternal RB1 allele that is mutated [51], presumably due to the increased exposure risks of sperms relative to ova. This observation is consistent with the epidemiologic data suggesting an increased risk for fathers in metal works and the military to have children with retinoblastoma [52,53].

2.Mosaicism of RB1 Mutations

About 10% of new RB1 germline mutations may be present in only a fraction of the cells of the proband. These individuals, therefore, are considered to be mosaic for the mutant allele. Mosaicism becomes obvious only when a parent with more than one child with retinoblastoma does not show the same mutant RB1 allele present in the children [54,55]. Of the 10% bilaterally affected probands in whom we cannot identify the RB1 mutant allele, perhaps 1% may be mosaic, such that the fraction of mutant cells in their blood leukocytes is insufficient for detection of the mutation. Therefore, when the two RB1 mutant alleles are characterized in the retinoblastoma tumor tissue but cannot be found in the blood DNA of a unilaterally affected proband, we report that there remains a 2–3% chance that the child is mosaic. Since mosaicism could not be inherited, ancestors and siblings are not considered at risk for retinoblastoma. However, each future offspring of such a proband should be checked for the two mutant alleles found in the retinoblastoma tumor.

D.Phenotype-Genotype Correlations

1.RB1 ‘‘Null’’ Mutations and Penetrance

The vast majority of RB1 mutant alleles are ‘‘null,’’ resulting in premature truncation of translation, unstable mRNA [42] and no detectable pRB protein, and predispose to bilateral retinoblastoma tumors in close to 100% of children. Of the probands with null germline mutations, 92% had bilateral tumors, while only 50% of probands with in-frame mutations developed bilateral tumors. The disease/eye ratio (DER), or ratio of clinically affected eyes to mutation carriers [56], for persons with null RB1 alleles is approximately 2.0 [57]. RB1 mutations that result in a stable but less active protein are associated with reduced penetrance and a reduced number of tumors (30% of eyes affected) versus 100% of eyes in persons with null alleles. The recurrent missense mutation R661W has been associated with reduced penetrance with an average DER of 0.73 [58,59]. The R661W protein has been shown to have partial activity, including retention of nuclear localization and hyperphosphorylation, but reduced or absent pRB-binding protein interactions [57,60,61]. Sporadic unilateral patients with R661W germline mutations are founders of reduced penetrance families. Other low penetrance missense mutations affect the promoter, the A/B domain of pRB, or splicing functions, resulting in a reduction of the amount of pRB [62–64].

2.Examples of Low-Penetrance Families

We reported a low penetrance family with an in-frame deletion of RB1 exon 24 and 25, resulting in a protein with altered nuclear localization, reduced ability to repress E2F-mediated transcription, inability to bind the MDM2 pRB-binding protein, and the inability to suppress growth [65]. Of 18 members of this large family known to carry the D24–25 mutation, only 11 of 18 individuals developed retinoblastoma, 1 had retinoma, and only 3 had bilateral tumors (low expressivity). The DER for this family was 0.78, consistent with reduced penetrance. Other small deletions that resulted in low penetrance retinoblastoma include DN480 that deletes asparagine 480, and D4 that contains an in-frame deletion of exon 4. Each of these mutants retained the ability to suppress growth but differed in their ability to bind pRBbinding proteins [57].

3.Other Primary Tumors from pRB Inactivation

Inactivation of pRB and other proteins in the pRB cell cycle regulatory pathway contribute to a wide array of cancers including osteosarcoma, small cell lung carcinoma, breast carcinoma, and other malignancies [66]. Inactivation of the pathway can occur by mutation of RB1, mutation of p16 which inhibits pRB phosphorylation by inhibiting cyclin D1, or amplification of cyclin D1, which phosphorylates pRB to its inactive form.

4.Exogenous Inactivation of pRB

Some of the missing RB1 mutations in retinoblastoma tumors theoretically may involve other parts of the RB pathway [67]. It has also been suggested that exogenous factors may induce retinoblastoma, since the DNA of the human papillomavirus has been detected in 36% of retinoblastoma tumors [68]. This virus is well known to produce proteins that can inactivate pRB. However, more work is needed to clarify these observations, since the RB1 mutations have not been studied in this particular set of tumors.

III.CANDIDATE M3-TO-MN EVENTS

A.Genomic Gains and Losses

1.Gains at 1q and 2p and Loss at 16q as M3-to-Mn Events

All retinoblastoma tumors are initiated by mutation of both copies of the RB1 gene. There is no evidence that any other gene besides RB1 initiates the predisposition to human retinoblastoma. Knudson’s two-hit hypothesis states that these two mutational events (M1 and M2) are rate-limiting for development of retinoblastoma [69]. However, additional mutations (M3-Mn) are almost certainly required before full malignant transformation results, since there are genetic alterations in addition to RB1 in all retinoblastoma tumors [4,5]. Karyotypic analysis and comparative genomic hybridization (CGH) have been used to characterize genomic gains and losses within retinoblastoma tumors [70–72]. The minimal regions most frequently gained were 1q31 (52%), 6p22 (44%), 2p24–25 (30%), and 13q32–34 (12%) [72]. The

minimal region most frequently lost was 16q22 (14%). In one study, gains at 1q and 2p and loss of 16q were restricted to more advanced tumors in older children, suggesting that these changes might correlate with tumor progression. However, 6p gains might be associated more with tumor initiation [71]. Chromosome 1q gain is common in many cancers besides retinoblastoma [73–77]. However, almost unique to retinoblastoma is a specific pattern of 6p gain, i(6p), identified in 60% of retinoblastoma tumors [78]. This marker chromosome results in four copies of genes on chromosome 6p, or low-level amplification [79].

2.Maintenance of Genetic Stability by pRB

Interestingly, recent data from the mouse suggests that pRB is involved in maintaining genetic stability within the cell. Using a retrovirus carrying negative and positive selectable markers which integrated randomly into individual chromosomes, the authors were able to determine the frequency of loss of the chromosomal marker in mammalian cells [80]. In normal mouse embryonic stem (ES) cells, the frequency for loss of the marker was less than 10 8 per cell per generation. In RB1 / ES cells, the frequency was increased to 10 5 per cell per generation, while in RB1þ/ ES cells, the frequency was 10 7 per cell per generation. Such a process may account for the emergence of M3 events after the loss of RB1. However, in contrast to adult-onset tumors, retinoblastoma tumors have abnormal karyotypes with a finite number of genomic changes that remain stable over years.

B.Isochromosome 6p Contains a Candidate Oncogene, RBKIN

We hypothesized that the region of chromosome 6p gain may contain an oncogene necessary for progression of retinoblastoma. To identify candidate oncogenes, we narrowed the region of gain on 6p to a minimal 0.6-Mb region of 6p22 using quantitative multiplex polymerase chain reaction (QM-PCR) for sequence-tagged sites (STS) [81], and cloned a novel human kinesin-like gene, RBKIN. RBKIN expression is increased in retinoblastoma compared to normal human retina, and inhibition of RBKIN in retinoblastoma cell lines using an antisense oligonucleotide results in a block in cell growth, consistent with RBKIN acting as an oncogene [81]. However, further experimentation is necessary to confirm the role of RBKIN as an oncogene in retinoblastoma development. Chromosome 6p gain is also common in bladder cancer and is associated with an elevated risk of progression of bladder cancer [82,83], suggesting that RBKIN may have a much broader role in cancer.

C.Chromosome 16q Loss Narrowed to a Member of the Cadherin Family

Using loss of heterozygosity and QM-PCR, we narrowed the loss of chromosome 16q22 in retinoblastoma to a region that contains the cadherin 11 (CDH11) and CDH13 genes. Allelic loss on 16q22 has been frequently shown in liver [84], breast [85], prostate [86], and Wilms’ [87] tumors, suggesting the presence of a gene or genes in this region that may be important in carcinogenesis. Three members of the cadherin family have previously been implicated in carcinogenesis. Downregulation of CDH1 gene expression correlates with invasive potential and poor prognosis in

human carcinoma, sometimes combined with functional loss of CDH11 [88]. CDH11 is also downregulated in 23 cases of osteosarcoma but not in normal osteoblasts, and in astrocytoma but not in normal brain tissues [89,90]. The CDH13 gene has been deleted or inactivated by promoter methylation in more than 50% of advanced-stage ovarian cancers [91] and in breast, lung, and osteosarcoma tumors [92,93]. We have shown the expression of CDH11 in normal retina, but not in 2 of 3 primary retinoblastoma tumors and 3 of 4 retinoblastoma cell lines. One primary retinoblastoma and one cell line showed a faster migrating form of CDH11. The role of cadherins in progression of retinoblastoma requires further investigation.

IV. SLIPPING PAST CELL DEATH

A.Importance of Programmed Cell Death in Retinal Development

The retina is one of the tissues whose final architecture is achieved by the precise balance between proliferation, differentiation, and apoptosis. During retinal development, the number of cells produced exceeds the number of cells ultimately required, so that many cells are removed by apoptosis to achieve the final retinal structure.

1.Role of p75NTR in the Retina and in Retinoblastoma

Nerve growth factor (NGF) signals for survival through TrkA, and for apoptosis through p75NTR in retina [94–96]. Induction of cell death through p75NTR may be preceded by the unscheduled re-entry of postmitotic neurons into the cell cycle [95]. We have evidence that downregulation of the p75NTR protein might be important for M3-Mn events in retinoblastoma by promoting proliferation of RB1 / retinal cells and disrupting apoptosis induced by NGF produced by Mu¨ller glia.

The p75NTR protein is highly expressed in Mu¨ller glia of wild-type retina and in the benign retinoma but is undetectable in retinoblastoma (unpublished data). Although we found significant cell division in retinoma, indicated by expression of proliferating cell nuclear antigen (PCNA), there is probably also significant apoptotic activity in retinoma compared to retinoblastoma. Therefore there may be a homeostatic balance between cell growth and cell death within a retinoma, explaining why retinomas are generally stable. One or more additional mutations that prevent apoptosis may tip over this balance and convert a retinoma into a retinoblastoma.

B.Telomerase and Immortality

1.Telomerase Activity in Cancers

The role of telomerase in the development of cancer has been suggested based on the presence of telomerase activity in many human cancers, whereas normal cells are devoid of telomerase activity [97]. Telomeres are specialized protein-DNA structures at the ends of chromosomes that protect the chromosome from end-to-end fusion and eventual cell death [98]. Telomerase is the reverse transcriptase that maintains the ends of chromosomes. During normal cellular replication, the lagging strand is

replicated by short RNA primers, which are made by RNA primase. These primers are extended by DNA polymerase to form Okazaki fragments [99]. However, when these primers are removed, there is no way to replicate the small fragment of DNA at the ends of the chromosomes. Therefore, with each cell division, the chromosomes are shortened a bit more and eventually become inadequate, causing the cell to enter crisis. In order for a cell to become immortal, it must overcome the end replication problem and maintain intact telomeres. Cancer cells are thought to do this by reactivating telomerase, which has been found to be highly expressed in a number of cancers, including metastatic ovarian carcinoma, colorectal adenocarcinoma, and aggressively proliferating myeloid leukemia [100–103].

However, few retinoblastoma tumors activate telomerase en route to malignancy. Of 34 retinoblastoma tumors studied, 17 (50%) did not express telomerase [104]. Telomerase activity even in the 17 telomerase-positive retinoblastoma tumors was low relative to the telomerase-positive adenovirus-transformed 293 cells. As expected, normal human embryonic retina was negative for telomerase activity. Unilateral and bilateral tumors were almost equally represented in the telomerase-positive and telomerase-negative groups. Additionally, those telomerasenegative tumors contained significantly longer telomeres than the telomerasepositive tumors. Telomerase-negative retinoblastoma tumors containing longer telomeres may have acquired mutations that utilize different mechanisms, such as activation of an oncogene or loss of a tumor suppressor gene, thereby bypassing the need for reactivation of telomerase.

V.RETINOBLASTOMA RESISTANCE TO THERAPY

Chemotherapy followed by focal therapy with laser and cryotherapy has become the primary treatment for large intraocular retinoblastomas and visually threatening small tumors. Extraocular retinoblastoma can also respond to chemotherapy, and in combination with orbital radiation and bone marrow or stem cell transplantation, there is the possibility of long-term remissions and even cures of extraocular retinoblastoma [105,106]. A large series treated with a unified protocol has yet to be reported.

A.Multidrug-Resistance Genes

1.Overview of Multidrug Resistance in Cancer

Retinoblastoma tumors are rarely cured by chemotherapy alone, since they frequently contain cells that are resistant and survive chemotherapy to regrow. Multidrug resistance (MDR) in the human host could be caused by a number of genes and factors acting cooperatively (Fig. 3). Upstream factors include genetic factors intrinsic to the host, tumor factors, drug metabolism, drug clearance, and drug distribution, all potentially affecting the response to chemotherapy. Tumor factors include a variety of genes that confer multidrug resistance, such as MDR1 (P- glycoprotein), MRP (multidrug-resistance protein), LRP (lung resistance protein), Topo II (topoisomerase II enzymes), and GSH (glutathione enzymes), all potentially affecting the response to chemotherapy. Downstream factors include yet other genes

Figure 3 Model of multidrug resistance and mechanism of cyclosporine (CSA) action. A. Chemotherapy drugs diffuse into the tumor cell but are rapidly removed by the MDR membrane pump, allowing tumor cell survival. Some cells die, but many are resistant. Multidrug resistance correlates with the level of expression of the MDR protein (shown using immunohistochemistry), so residual and recurrent retinoblastomas stain positive for MDR and negative for MRP (not shown). B. CSA given at the same time as the chemotherapy drugs is removed from the tumor cell by the MDR membrane pump, competing with extrusion of chemotherapy drugs. The chemotherapy drugs remain in the tumor cell, achieving a high rate of cell death. Retinoblastoma multidrug resistance is less likely; but when it arises, the recurrent cells are MDR-negative and MRP-positive (not shown).

that regulate cell cycle progression, cell death, and differentiation, including the bcl-2 family of apoptosis genes, oncogenes (e.g., myc family genes), and tumor suppressor genes (e.g., p53 family genes) [107,108].

2.Classic Multidrug-Resistance P-Glycoprotein in Cancer

Ling and coworkers first described the classic MDR phenotype in human cancer, which they have shown to be due to upregulation of the MDR1 gene, with increased expression of P-glycoprotein [109]. This MDR protein broadly reduces intracellular levels of vinca alkaloids (vincristine, vinblastine), epipodophyllotoxins (etoposide, teniposide) and other natural-product antineoplastic agents, by functioning as an ATP-dependent plasma membrane pump that expels drugs from human cancer cells

[110] (Fig. 3). Platinum compounds are not substrates of the P-glycoprotein drug efflux mechanism.

Relatively high expression of P-glycoprotein is found in normal epithelial tissues and in cancers derived from epithelial tissues arising from the adrenal cortex, breast, kidney, liver, ovary and colon [111–114]. Increased expression of P- glycoprotein has been reported in leukemia, lymphoma and myeloma, neuroblastoma, osteosarcoma, rhabdomyosarcoma, and retinoblastoma [115–128]. Increased P-glycoprotein at diagnosis in these cancers generally correlates with failure of therapy, whereas undetectable P-glycoprotein at diagnosis correlates with lasting remission [115–131].

The increased P-glycoprotein that we found in many retinoblastoma tumors when the eye was removed for relapse may have accounted for the frequent chemotherapy failures observed in the past 30 years [122,126,128]. Likewise, the good response of retinoblastoma to carboplatin, teniposide, and vincristine that we have reported might be due to concurrent use of cyclosporin A (CSA) [132,133], a known inhibitor of P-glycoprotein in vitro, that inhibits the efflux of drugs by the MDR pump, allowing the drugs to be effective [128].

For intraocular retinoblastoma, it is not possible to correlate P-glycoprotein levels before therapy with the outcome of chemotherapy since biopsies are known to incur an increased risk of systemic spread [134]. In metastatic retinoblastoma, however, P-glycoprotein expression before therapy correlates precisely with failure of therapy (Fig. 3), and initially undetectable P-glycoprotein correlates with longterm remission [122,126,128]. We have found that P-glycoprotein is increased in onethird of eyes with large retinoblastoma enucleated primarily at diagnosis and is present in all eyes with large retinoblastoma enucleated at failure of different primary therapies. This suggests that the intrinsic presence of P-glycoprotein before therapy and/or its subsequent induction by therapy may both be responsible for the resistance of large retinoblastoma to chemotherapy [122,126,128]. Eyes with medium and small retinoblastoma often respond well to chemotherapy, but they almost always require focal therapy for residual or recurrent disease [135–139].

In retinoblastoma cell lines, the MDR phenotype correlated with increased P- glycoprotein in the original retinoblastoma tumors that were resistant to chemotherapy in vitro [122,126,128,140] (Fig. 3). Conversely, P-glycoprotein was undetectable in retinoblastoma cell lines and the original retinoblastoma tumors that appeared sensitive to chemotherapy in vitro [122,126,128,140]. Poor penetration of chemotherapy into the eye might also contribute to drug resistance in retinoblastoma. The use of CSA doubled the intravitreal carboplatin levels in animals [141], possibly by circumventing the effect of the highly expressed P-glycoprotein at the ‘‘blood-eye barrier’’ [142–144]. The mechanism has not been determined for this observation.

3.Other Multidrug-Resistance Genes Expressed in Retinoblastoma

Intraocular retinoblastoma that failed chemotherapy despite CSA use expressed another multidrug-resistance protein (MRP) rather than P-glycoprotein, whereas retinoblastoma that failed chemotherapy prior to the use of CSA showed increased P-glycoprotein [128]. MRP belongs to the same ATP-dependent membrane transporter superfamily as P-glycoprotein [145], and transfection of the MRP gene

confers a similar broad-spectrum pattern of drug resistance to antineoplastic agents [122,146]. Whereas CSA inhibits P-glycoprotein in vitro, CSA does not block MRP, for which there is presently no effective inhibitor. We have shown that the presence of MRP in retinoblastoma may be associated with failure despite using CSA concurrently with chemotherapy [128]. However, MRP is less frequently found at diagnosis in retinoblastoma (1 of 18) than P-glycoprotein (8 of 38).

B.Modification of Therapy to Avoid Multidrug Resistance

Eye irradiation of bilateral patients with RB1 germline mutations and large intraocular tumors results in an at least 30% risk of secondary bone or soft tissue sarcoma, brain tumor, lung tumor, or malignant melanoma by 30 years [67,148– 151] and a 90% risk of orbital deformities and lacrimal dysfunction [152,153]. To avoid radiation-induced secondary tumors, chemotherapy has become the primary treatment in most centers. To circumvent multidrug resistance, we have added CSA to block the P-glycoprotein-induced drug efflux (Fig. 3). CSA and chemotherapy (vincristine and etoposide or teniposide with or without carboplatin) controls intraocular retinoblastoma without requiring radiation: 91% of newly untreated tumors remained relapse-free, and 70% of tumors that had relapsed from a previous therapy were cured [132,133]. In addition, tumors with the worst prognosis (vitreous seeds) were 88% relapse-free, better than previously reported [132,133].

The Toronto retinoblastoma protocol with short 3-hr infusions of high-dose CSA showed increased efficacy but no concurrent increase in chemotherapy toxicity, as is observed with prolonged infusions of CSA [154–158]. Besides reversing MDR, preclinical data suggest that CSA might increase the efficacy of carboplatin by modulating non-P-glycoprotein mechanisms of drug resistance. CSA given in short high doses may also prevent P-glycoprotein upregulation by chemotherapy agents [159,160]. A recent in vitro study also suggests that CSA might induce cancer progression by a cell-autonomous mechanism [161]. Therefore we will continue to cautiously use CSA for modulation of chemotherapy, because our own clinical data strongly suggest that CSA improves the long-term response of retinoblastoma to chemotherapy [162].

VI. THERAPY-INDUCED MUTATIONS LEAD TO SECOND PRIMARY TUMORS IN RETINOBLASTOMA PATIENTS

A.Radiation Therapy

The radiation-induced tumors are aneuploid, with many chromosomal markers and genetic rearrangements. Presumably, the susceptibility of RB1þ/ persons to second primary tumors is due to the mutation of the second RB1 allele induced by radiation.