Rapid Identification of Germline Mutations in Retinoblastoma by ProteinTruncation Testing FREE

Retinoblastoma (RB) affects approximately 1 in 20 000 children^{1 - 2} and accounts for 12% of infant cancersin the United States.³ The disease arises fromloss or mutation of both alleles of the RB tumor suppressor gene (RB1). Retinoblastoma is divided into 2 types, heritable and nonheritable,according to the mechanism by which the 2 RB1 allelesare inactivated. In nonheritable RB, both alleles are inactivated somaticallywithin a single retinal cell, resulting in a unilateral, unifocal tumor. Inheritable RB, 1 RB1 allele is inactivated in thegermline and loss or mutation of the second allele occurs somatically. Childrenwith heritable RB typically develop bilateral, multifocal retinal tumors,although 15% of children with unilateral disease also harbor an underlyinggermline mutation.^{4 - 5}

The distinction between heritable and nonheritable RB is critical fortreatment planning and disease management. Children with heritable RB, inaddition to experiencing bilateral vision impairment in early childhood, areat increased risk for developing primitive neuroectodermal tumors and secondtumors later in life. While rare, primitive neuroectodermal tumors representa significant threat because they are uniformly and rapidly fatal. Other tumorsalso represent a significant problem, with 50% of patients with heritableRB who survive childhood succumbing to second tumors within 50 years afterdiagnosis.⁶ Additionally, patients with heritableRB have a nearly 50% chance of passing this genetic defect on to their children.^{7 - 8} Heritable and nonheritable RB casesare not always clinically distinguishable, especially in patients with unilateraldisease. In many cases, genetic testing for the presence of a germline RB1 mutation is the only means by which the diagnosis maybe established with certainty.

Direct genetic testing for germline mutations that underlie heritableRB first became possible when RB1 was identifiedin 1986.^{9 - 11} Despiteknowledge of the complete genomic sequence, 3 obstacles have made efficient RB1 genetic testing difficult: (1) RB1 is large, containing 27 exons that span approximately 180 000bases.^{9 - 11} (2)No known mutation hot spots have emerged, despite the description of morethan 230 different germline RB1 mutations.^{12 - 13} (3) The majority of mutations arisefrom small (< 10 base pair [bp]) sequence changes. These obstacles havetraditionally necessitated detailed screening of all 27 RB1 exons and adjacent introns to locate a tumorigenic alteration.This kind of testing is costly, and the time required for obtaining a positiveresult often reduces clinical utility.

Analysis of published RB1 mutations, however,reveals that despite broad variation in mutation location and type, up to90% of all RB1 mutations result in a premature stopcodon.¹² This commonality is well suited todetection by protein truncation testing (PTT) using in vitro transcriptionand translation (IVTT).¹⁴ In PTT, the in vitrosynthesis of protein from amplified RNA screens the coding region of a genefor mutations that result in protein truncation.

Because PTT can rapidly detect truncating mutations from a peripheralblood sample, it has been applied with success in detecting mutations in genesresponsible for several heritable forms of cancer, including colon cancer(FAP, HNPCC), breast cancer(BRCA1, BRCA2), tuberoussclerosis (TSC1, TSC2),and neurofibromatosis (NF1, NF2).¹⁵ As in RB1,the vast majority of the reported mutations in these genes result in a prematuretranslation termination. One research group has applied PTT to detect RB1 mutations in tumor specimens. In that study, positiveresults were obtained in 5 of 35 malignant fibrous histiocytoma tumor samples(14%).¹⁶ However, PTT has yet to be appliedto detect germline RB1 mutations in peripheral bloodsamples. Peripheral blood analysis has particular clinical utility in RB becausethese patients are often treated with ocular conservation and their tumorspecimens are not readily available for genetic analysis.

Once a truncating mutation has been detected by PTT, the relative degreeof protein truncation present pinpoints the site of the underlying mutation.This localizing information has the benefit of expediting mutation identificationby obviating the need to screen the entire RB1 gene.Only a small fraction of the gene, determined by the results of PTT, requiressequencing. This benefit of PTT produces savings in time and effort requiredfor mutation detection and provides genetic information in a time frame thatis useful for patient counseling and for clinical decision making. With theseadvantages in mind, an RB1 genetic testing protocolusing PTT was developed for clinical testing in the Ocular Oncology Unit atthe University of California, San Francisco.

Twenty-seven patients with the clinical diagnosis of hereditary RB from27 separate kindreds were enrolled with written informed consent as probandsfor RB1 genetic testing by PTT. All probands wereestablished patients in the clinical practice of one ocular oncologist (J.O. B.) and were enrolled consecutively at the time of their regularly scheduledvisits. All procedures, protocols, and consent forms were approved by theUniversity of California, San Francisco, Committee on Human Research and followedthe tenets of the Declaration of Helsinki. The clinical diagnosis of hereditaryRB was determined by the presence of bilateral tumors, multifocal unilateraltumors, or a unilateral tumor concurrent with a positive family history ofRB. Additionally, 13 relatives from 7 of the 27 kindreds were enrolled forfamilial testing. Seven of these family members were unaffected by RB; theremaining 6 had either unilateral or bilateral RB tumors. Ten unrelated patientswith neither a personal nor a family history of RB were enrolled as negativecontrol subjects.

To maximize the advantages of PTT and to minimize the number of patientsrequiring full RB1 screening, 1 affected member fromeach kindred was selected as the proband for PTT. If PTT detected a germlinemutation in this proband, targeted complementary DNA (cDNA) and genomic sequencingwere undertaken to identify the precise DNA alteration. When a mutation wasidentified, direct testing of available relatives by targeted genomic sequencingwas undertaken to determine if the proband's mutation was shared. If PTT didnot detect a mutation in the proband, the test was repeated on a new bloodsample from the same patient. Additionally, when available, another affectedmember of the kindred was selected as a second subject for PTT. If no secondaffected family member was available or if both affected family members hadnegative PTT results, then exons 2 to 27 and the promoter region were systematicallysequenced from the proband's polymerase chain reaction (PCR)–amplifiedgenomic DNA. When possible, a new blood sample was collected from patientsdisplaying negative results using PTT. Cytogenetic analysis by Giemsa-trypsin-Giemsa(GTG) banding and by fluorescent in situ hybridization (FISH) for RB1 was performed.

The general scheme for PTT is illustrated in Figure 1.

Within 2 hours after collection in heparin sodium, blood samples weretransported to the laboratory at room temperature and incubated with gentleinversion for 4 hours at 37°C in 200 µg/mL of puromycin. Total RNAwas then isolated from leukocytes using QIAamp RNA Blood Mini Kits (QIAGEN,Valencia, Calif) according to the manufacturer's protocol. Synthesis of first-strandcDNA was performed on 0.5 to 1.0 µg of total RNA with the SuperscriptFirst-Strand Synthesis System for RT-PCR (Gibco-BRL, Rockville, Md) usinga primer mixture containing oligo dT and 2 RB1-specificantisense primers, RbRT (5′-GACTAACATTTCAAGTGGC-3′), which primes128 bp downstream of the endogenous stop codon, and RbRT2 (5′-GAGGTAGATTTCAATGGCT-3′),which primes at amino acid 646 within the wild-type open reading frame. Reactionswere performed at 42°C for 90 minutes.

The PCR amplification of the RB1 coding regionwas performed in 3 overlapping fragments. Each 50 µL reaction contained3 to 5 µL of first-strand cDNA; 10 pmol each of the sense and antisenseprimers (Table 1); 10 µLof 5× Q-Solution (QIAGEN); 5 µL of 10× PCR buffer (QIAGEN);4.0mM of magnesium chloride; 0.2mM each of dATP, dCTP, dGTP, and dTTP; and2 units of Taq polymerase (QIAGEN). Sense primers contained the T7 promotersequence, a eukaryotic translation initiation sequence (consensus Kozak sequence),and an N-terminal myc-tag, MEQKLISEEDLN (Figure 2).¹⁷ Reactions were performedwith an initial denaturation at 94°C for 3 minutes with 35 cycles of 94°Cfor 1 minute, 55°C for 1 minute, and 72°C for 3 minutes, followedby a final polymerization at 72°C for 7 minutes. We confirmed the reversetranscription PCR (RT-PCR) success by visualizing 5 µL of the PCR reactionwith ethidium bromide staining after electrophoresis through a 0.8% agarosegel. Large DNA sequence deletions or insertions couldbe detected at this point by comparison with wild-type RT-PCR products.

We performed IVTT with the TNT T7 Quick Coupled Transcription/TranslationSystem (Promega, Madison, Wis) at 30°C for 90 minutes. Each 25 µLreaction contained 4 µL of RT-PCR product and 10 µCi (370 000Bq) of ³⁵S-methionine (Amersham, Piscataway, NJ). The IVTT productswere immunoprecipitated by adding 10 µg of agarose-conjugated anti-c-myc(9E10) mouse Mab (Santa Cruz Biotechnology, Santa Cruz, Calif) and 1 mL ofphosphate-buffered saline. After incubation with gentle inversion at 4°Cfor more than 2 hours, samples were pelleted and washed 4 times with phosphate-bufferedsaline.

Electrophoresis was performed through 12% sodium dodecyl sulfate–polyacrylamidegels after resuspension of the immunoprecipitated pellet in Laemmli loadingbuffer (Bio-Rad, Hercules, Calif) and denaturation at 95°C for 5 minutes.Autoradiographic visualization was performed after gels were fixed for 15minutes in acetic acid–methonol-water (10:50:40), soaked for 15 minutesin Amplify fluorographic reagent (Amersham), vacuum dried at 65°C for90 minutes, and exposed to film at −80°C for 1 to 4 days. A mutationwas detected when visual comparison revealed a patient protein size smallerthan that of wild-type proteins.

Templates for cDNA sequencing were generated by RT-PCR, using the sameprimers and conditions used in PTT (Table1). These sequencing templates were purified from free nucleotidesand primers using the QIAquick PCR Purification Kit (QIAGEN). Then, 2 to 4overlapping primers from a set of 18 spanning the entire RB1 coding region (Table 2)were selected to target direct sequencing to the region most likely to harborthe truncating mutation, based on the size of the mutant protein relativeto the size of the wild-type protein. Sequencing was performed on an ABI Prism377 DNA sequencer (PE Biosystems, Foster City, Calif). Electropherograms ofsequenced regions were compared visually with wild-type electropherogramsfor the presence of sequence anomalies resulting in premature stop codons.All cDNA mutations were confirmed by genomic DNA sequencing.

Concurrent with the blood draw for PTT, whole blood was collected frompatients for genomic DNA preparation and stored at −20°C. For genomicDNA isolation, 400 µL of this blood was thawed and processed using theQIAamp Blood Kit (QIAGEN) according to the manufacturer's protocol. Sequencingtemplates were generated from genomic DNA by PCR amplification of RB1 exons and their adjacent intronic sequences (Table 3). Specific exons were selected to direct sequencing to theregion most likely to contain the premature stop codon based on either theresults of cDNA sequencing or the size of truncated PTT products. For patientswith hereditary RB in whom PTT results were negative, exons 2 to 27 and thepromoter region (including −327 to −89)¹⁸ wereamplified for sequencing. Exon 1 was not included because of difficultieswith amplification in this guanine- and cytosine-rich region of RB1. Amplification products were purified from free nucleotides andprimers using the QIAquick PCR Purification Kit (QIAGEN) according to themanufacturer's protocol. For most exons and adjacent intronic sequences, thesame primers used for PCR amplification were used to sequence both strandsof the purified templates. Exceptions are noted in Table 3.

Sequencing was performed on an ABI Prism 377 DNA sequencer (PE Biosystems).Electropherograms of sequenced exons and adjacent introns were compared visuallywith wild-type electropherograms for the presence of sequence anomalies. Allmutations were confirmed by sequencing both strands of genomic DNA, by correlationwith PTT truncated fragment size, and by repeat sequencing from a second genomicDNA preparation from the same patient. The RB1 referencesequence corresponded to GenBank accession number L11910.

When an additional blood sample was available for patients in whom PTTresults were negative, cytogenetic analysis by GTG banding and by FISH for RB1 were performed according to protocols of the CytogeneticsLaboratory at the Children's Hospital of Oakland, Oakland, Calif.

Of the 27 probands tested, 19 (70%) tested positive for germline mutationsby PTT (Figure 3). Genomic DNA sequenceconfirmation on probands with positive PTT results revealed 3 patients withsmall frameshifting deletions, 6 with splice site mutations, and 9 with singlebase pair substitutions resulting in nonsense mutations (Figure 3 and Figure 4)(Table 4). The proband from the1 remaining kindred (Rb231) demonstrated an approximately 20 kDa truncatedprotein product with PTT of the most 3′ of the overlapping RT-PCR templates.This product would suggest an early termination codon near exon 18, but focusedsequencing of both cDNA and genomic DNA that included splice sites and codingregions for exons 17 to 20 failed to reveal the truncating mutation.

In 8 probands, the initial PTT results were negative. When repeatedon a second blood sample, all of these probands had negative results a secondtime. For 3 of these probands (Rb204, Rb209, and Rb222), a second family memberaffected by RB was available for PTT (Rb205, Rb210, and Rb232, respectively).Protein truncation testing results were also negative for 2 of these otheraffected family members (Rb205 and Rb210). One second family member (Rb232)demonstrated positive PTT results that were subsequently identified as resultingfrom a frameshifting mutation (Table 4).

In all, 10 patients with hereditary RB had negative PTT results. Eightwere probands and 2 were affected second family members in 2 of the 8 kindreds.All 10 of these were further analyzed by systematic sequencing of exons 2to 27 and the promoter region. Systematic sequencing identified 2 probandswith missense mutations (Rb177 and Rb222) and 1 proband (Rb219) with a splicesite mutation that would result in an inframe deletion. Three of the 10 patientswith negative PTT results, representing 3 different kindreds, had additionalblood drawn for cytogenetic analysis. Cytogenetic analysis by GTG bandingrevealed 1 proband (Rb207) with an interstitial deletion of 1 chromosome 13spanning RB1 (break points at 13q12 and 13q14) thatwas confirmed by FISH (Table 5).

The results of PTT were used for focused screening of 11 relatives in7 kindreds. Three kindreds had multiple family members with RB, and in all3, the affected family members shared a mutation with the affected proband.In 1 kindred with 2 affected siblings (Rb222 and Rb232), 2 germline mutationswere identified. One sibling (Rb232) had positive PTT results as a secondaffected family member, and a truncating mutation was revealed by direct sequencing,a 4 base pair deletion in exon 18 (g. 150029-150032del). This is a frameshiftmutation that produces a new stop codon in exon 19. Familial testing of theproband sibling found no mutation at this location. Systematic exon-by-exonsequencing revealed that both siblings shared a missense mutation (g.70243A>T),a single base change that causes the threonine at amino acid position 377to be substituted with serine.

None of the 7 unaffected family members tested shared the causativemutation identified in an affected relative. Of the 10 control patients, 9tested negative for germline mutations by PTT and 1 had an insufficient amountof total RNA isolated for analysis.

Despite the availability of many molecular diagnostic techniques forthe detection of germline mutations in RB1, clinicalapplication of this genetic testing has not been widespread. Available moleculartechniques such as Southern blotting, single-strand conformation polymorphismanalysis, heteroduplex analysis, restriction fragment length polymorphismanalysis, and direct sequencing can be prohibitively time-consuming or expensive.Modalities that are performed efficiently or are commercially available, suchas cytogenetic analysis, are able to detect only a small fraction of the totalgermline mutations that may be present in RB1. Asingle procedure is needed that can function as a rapid screen for a majorityof RB1 germline mutations.

An ideal test for germline mutations in RB1 wouldincorporate the following features: (1) a noninvasive method of sample collection,(2) the capacity to detect all mutation types that may exist in RB1, (3) specificity for RB1 sequence alterationsthat are pathogenic (neutral variations and polymorphisms would not be detected),(4) rapid and cost-effective performance, and (5) sensitivity sufficient toassure that all negative results are true negative results. While it is unlikelythat any single procedure can meet all of these criteria, to our knowledge,PTT fulfills more of these than any other genetic testing approach to date.

Samples for PTT are collected relatively noninvasively by phlebotomy.Because of its high sensitivity for truncating mutations, PTT requires onlya small fraction of leukocytes to carry the mutation in order to detect alterationsin patients with somatic mosaic RB. Since the sensitivity of PTT was 70% inprobands with hereditary RB, one could not safely exclude the presence ofa germline mutation in a patient with unilateral RB using PTT alone. In caseswhere no mutation is detected by PTT in the blood, primary analysis of tumortissue followed by a secondary search for tumor-detected mutations in bloodsamples would allow for more confident diagnoses of nonhereditary RB.

Although PTT is not sensitive to all mutation types, it detects a broadrange of the most common RB1 mutation types: smallframeshift mutations, splice site alterations, and nonsense mutations. Proteintruncation testing can also readily detect some rare mutation types that maybe difficult to detect with genomic screening alone, such as an intronic mutationaffecting a splice site adjacent to an exon. Genomic sequencing of exon andflanking intron DNA will miss deletions or insertions with break points deeperinto the intronic sequences that result in messenger RNA truncated by theabsence of 1 or more exons. While focused sequencing of cDNA and genomic DNAled to identification of truncating mutations in 19 (95%) of 20 kindreds,the presence of an internal intron mutation affecting splicing may explainwhy focused sequencing of genomic DNA did not identify a truncating mutationin the remaining kindred (Rb231).

Mutation types exist that PTT cannot detect, 2 of which, missense mutationsand large chromosomal deletions, were found in this study on further analysisof probands with negative PTT results. Other known RB1 mutationtypes not detectable by PTT and not discovered among the negative samplesin this study include promoter mutations, small in-frame deletions, and chromosomaltranslocations. Fortunately, the mutation types that evade detection by PTTrepresent a minority of RB1 germline mutations. Missense,promoter, and in-frame deletions have been estimated to represent 11% of RB1 germline mutations.¹² Withthe prevalence of chromosomal alterations in patients with bilateral RB estimatedat 7% to 8%,¹⁹ the remaining 80% of patientswith nonsense, frameshift, and splice site mutations represent a rich opportunityfor detecting a majority of RB1 germline mutationsusing PTT as a single modality.

In this study, 19 (70%) of 27 probands had a germline mutation detectableby PTT. By using PTT as an initial screen, one can be more selective and efficientin the use of labor-intensive tests for mutations undetectable by PTT. Evenon the small scale of this study, limiting additional testing to those whohad negative PTT results increased the yield of subsequent testing. Detailedmolecular analysis results for missense mutations by genomic PCR and systematicsequencing of exons 2 to 27 and the promoter region were positive for 3 (30%)of 10 probands. Additional blood samples for cytogenetic analysis by GTG bandingand FISH at the RB1 locus were positive for a largedeletion in 1 of 3 probands whose PTT results were negative. With the prevalenceof missense mutations and of cytogenetically detectable alterations estimatedat 6% and 7% to 8%, respectively,^{12 ,19} thedetection rates in this study are boosted by first eliminating those patientswho have PTT detectable mutations.

Many mutation detection schemes capable of detecting single base pairchanges will also detect silent nonpathogenic alterations, false-positiveresults that must be distinguished from more functionally relevant alterations.Protein truncation testing has the advantage of detecting only mutation typesthat result in protein truncation, and these mutations are generally pathogenic.The truncating mutations reported in this study varied between kindreds; however,16 (84%) of 19 mutations were predicted to disrupt the small pocket domain(Figure 5), and all were predictedto affect the nuclear localization signal contained in exon 25. These areregions of the RB protein that have demonstrated functional significance.^{20 - 22} No truncated RB1 protein products were detected in healthy control subjects.The absence of known truncating polymorphisms in RB1 supportsthe pathogenic likelihood of a positive PTT result.

The presence of missense and promoter mutations should always raisethe possibility that they may be nontumorigenic polymorphisms. In this study,2 missense mutations went undetected by PTT, as expected, but were discoveredon systematic DNA sequencing. One occurred at the site of a previously definedpolymorphism in exon 20.²³ Whereas the describedpolymorphism at position g.156713 (C, 95%; A, 5%) results in no amino acidchange (a silent polymorphism), the mutation identified in Rb177 (g.156713C>T) results in alteration of an arginine codon (CGG) into a tryptophan codon(TGG). In the other missense mutation (Rb 222; g.70234A>T), an A-to-T substitutionresults in the alteration of the codon for amino acid 377 in exon 12 fromthreonine to serine. Both missense mutations are presumed to be tumorigenic,but functional analysis has not yet been performed for confirmation.

Perhaps the greatest advantage of PTT in RB1 genetictesting is its rapidity in detecting a positive result. Using the protocolapplied in this study, a truncated protein product can be detected in as fewas 7 days. The detection of a truncated protein by PTT signals the presenceof a germline truncating mutation. This confirms the diagnosis of hereditaryRB, but sequencing must be performed to define the exact mutation. The scopeof the search is narrowed, however, because the size of the truncated proteinrelative to the wild-type product provides an estimate of the location ofthe premature stop codon. Therefore, sequencing either of cDNA or genomicDNA may be focused on a small segment of the RB1 gene.Positive PTT results in this study required 2 to 4 weeks for sequence confirmationwith the aid of automated DNA sequencing.

Sequence confirmation provides the power to rapidly test both affectedand unaffected relatives for a familial mutation. Testing of relatives isexpedited because this testing can be focused on the presence or absence ofthe specific mutation detected in the proband. The PCR amplification fromgenomic DNA followed by sequencing of a single exon is all that is required.In this study, the results of all familial testing were consistent with thesubject's clinical phenotype.

An unexpected result of familial testing is the discovery of a kindredin which both affected siblings share a missense mutation and 1 of these alsocarries a second truncating germline mutation (Rb222 and Rb232). Rarely, familiesaffected by RB have been known to have individuals with different germlinemutations.²⁴ The possibility exists that themissense mutation shared is a nonpathogenic polymorphism, that the diagnosisof hereditary RB is incorrect (although the histopathologic features of theenucleated specimen are consistent with this diagnosis), or that a differentpathogenic mutation shared by both affected siblings has gone undetected.Further genetic analysis of this kindred is in progress.

The cost savings of PTT lie in the reduced time for analysis, savingboth laboratory time and technician labor. Material and reagent costs havebeen estimated at $200 to $250 per proband and $30 to $40 for each familymember subsequently tested. This suggests that a 4-member family could betested for fewer than $400 and that the results could be available in 2 to4 weeks. This compares favorably with existing methods of genetic analysiscommercially available, where testing a 4-member family could cost $3000 to$7500 and take 1 to 6 months.²⁵ Published analysisof the relative costs of molecular and conventional RB screening shows that,while health care savings are present initially, savings increase geometricallyas kindreds with known RB1 mutations grow.²⁶

In summary, the advantage of PTT lies not in its ability to detect all RB1 germline mutations but rather in its capacity to detectthe majority of mutations noninvasively, rapidly, and at low cost. No singlegenetic testing method currently available can detect all RB1 mutation types. The most effective and efficient strategies todate have used a series of nested tests. These protocols, some of which haveused as many as 5 different testing modalities, have final sensitivities rangingup to 83%.²⁷ Protein truncation testing iswell suited as the first step in a more comprehensive testing protocol, becauseit delivers a rapid, single test sensitivity (70%) in the range previouslyprovided only by multiple-modality testing, thereby allowing reservation ofmore intensive techniques for difficult cases. With a scheme of nested testingusing PTT as the first line and limited exon-by-exon sequencing and karyotypingapplied to individuals with negative PTT results to increase the sensitivity,the overall sensitivity is raised to 85% (23 of 27 kindreds with germlinemutations identified).

Future implications of a more efficient method of genetic testing arebroad. The clinical application of PTT in RB will improve detection of germlineRB mutations, which will supply critical information for prognosis, treatmentplanning, follow-up care, and genetic counseling. In addition to providingan adjunct to the clinical diagnosis of germline RB, PTT may permit presymptomaticdiagnosis and enable in utero screening. Identifying infants who inherit thisdisease for early medical intervention could maximize their survival and visualpotential. Protein truncation testing will allow rapid expansion of the publishedRB mutation database, with potential benefits for understanding the molecularand genetic bases for variations in the phenotype of this disease. If specificgenetic features predispose patients to better or worse clinical outcomes,patients could be stratified early in their disease course to identify thosewho require more aggressive clinical intervention.

The RB1 gene was the first tumor suppressorgene to be identified almost 2 decades ago. Despite this breakthrough, manyfamilies afflicted with this disease have never benefited from genetic testing.Protein truncation testing analysis is a tool that could make genetic testingmore available, affordable, and practical. Undoubtedly, this screening techniquewill be replaced by more powerful approaches in future. Effective, efficientgenetic screening will eventually be applied to broad populations at riskfor cancer and for many other diseases with a genetic basis. Until that time,PTT will allow us to accumulate RB1 mutations andto correlate these with clinical information in an effort to better diagnoseand treat this pediatric cancer syndrome.

Corresponding author and reprints: Joan M. O'Brien, MD, Director,Ocular Oncology Unit, Department of Ophthalmology, University of CaliforniaSan Francisco, 10 Koret Way, Box 0730, San Francisco, CA 94143.

Submitted for publication January 7, 2003; final revision received July15, 2003; accepted August 21, 2003.

This study was supported by the Knights Templar Eye Foundation Inc,Chicago, Ill; the Giannini Foundation, San Francisco, Calif; That Man MaySee Inc, San Francisco; the Sand Hill Foundation, Menlo Park, Calif; a Physician-ScientistAward from Research to Prevent Blindness, New York, NY (Dr O'Brien); and grantEY13812 (Dr O'Brien) and core grant EY02162 from the National Eye Institute,Bethesda, Md.

We would like to thank Anny Shai, BSc, for expert technical assistancein automated DNA sequencing.