Helen Dimaras1, Tim W. Corson2, David Cobrinik3, Abby White4, Junyang Zhao5, Francis L

Retinoblastoma

Helen Dimaras1, Tim W. Corson2, David Cobrinik3, Abby White4, Junyang Zhao5, Francis L. Munier6, David H. Abramson7, Carol L. Shields8, Guillermo L. Chantada9, Festus Njuguna10 and Brenda L. Gallie11

1| Department of Ophthalmology & Vision Sciences, The Hospital for Sick Children& University of Toronto, Toronto, Canada

2| Eugene and Marilyn Glick Eye Institute, Departments of Ophthalmology, Biochemistry and Molecular Biology, and Pharmacology and Toxicology, and Simon Cancer Center,Indiana University, Indianapolis, USA

3| The Vision Center and The Saban Research Institute Children's Hospital Los Angeles, Los Angeles, CA USA; and Department of Ophthalmology, Department of Biochemistry and Molecular Biology, The USC Eye Institute, and Norris Comprehensive Cancer Center, Keck School of Medicine of the University of Southern California, Los Angeles, CA USA

4| Daisy's Eye Cancer Fund, Oxford, UK

5| Department of Ophthalmology, Beijing Children’s Hospital, Capital Medial University, Beijing, China

6| Department of Ophthalmology, Jules-Gonin Eye Hospital, Lausanne, Switzerland

7| Department of Ophthalmology, Memorial Sloan Kettering Cancer Center, New York, USA

8| Ocular Oncology Service, Wills Eye Hospital, Philadelphia, USA

9| Hospital JP Garrahan, Buenos Aires, Argentina

10| Department of Pediatric Oncology, Moi Teaching & Referral Hospital, Eldoret, Kenya

11| Department of Ophthalmology & Vision Sciences, The Hospital for Sick Children& University of Toronto, 555 University Ave, Toronto, Ontario M5G1X8, Canada

Abstract

The genetic basis of human malignancy was determined through the study of retinoblastoma, a rare cancer of infant retina. Tumours form when both RB1 alleles mutate in a susceptible retinal cell, likely a cone photoreceptor precursor. The tumour suppressor functions of the retinoblastoma protein,pRB, relate to cell division and genomic stability, but the key biochemicalandmolecular basis oftissue specificity remain unknown. Retinoblastoma is diagnosed in8,000 children each year worldwide. Patientsurvivalis >95% in high-income countries, but <30% globally, depending on the socio-economical context. Stakeholder collaboration isimproving outcomes by increasing awareness for earlier diagnosis, sharing expertise, and developing guidelines. Intra-arterial and intra-vitreal chemotherapy have emerged as promising methods to salvage eyes.Ongoing international collaborationwill replace the multiple different classifications of eye involvement with standardized definitionsthat will facilitate assessment of eligibility, efficacy and safety of treatments. Heritable retinoblastoma survivors are at risk for second cancers; defining the molecular basis of RB1 retinal specificity may explain tissue specificity of second cancers, clearing the path to cancer prevention.

Keywords

retinoblastoma, pediatric oncology, pediatric ophthalmology, tumour suppressor gene, RB1 gene, MYCN oncogene, systemic chemotherapy, intra-arterial chemotherapy, intra-vitreal chemotherapy, cone photoreceptor, cancer therapy, global health, health equity

Competing interests

There is NO competing interest.

[H1] Introduction(BLG)

Retinoblastoma is a rare cancer initiated bybiallelicmutation of the retinoblastoma gene (RB1) in a single susceptible developing retinal cell. Inheritance of one mutant RB1 allele strongly predisposes to retinoblastomatumours that form when the second RB1 allele is mutated.1 Although RB1 loss initiates cancer in specific susceptible tissues, it is lost with progression in almost all human cancers.2The principles discovered by study of retinoblastoma led to the recognition that altered genes broadly underlie cancerinitiation and progression.

Retinoblastomastarts when the second RB1 allele is damaged in susceptible retinal cell (RB1-/-)that undergoes limited proliferation to form a non-malignant retinoma(Figure 1).3An intra-retinal tumourdevelops after genomic changes lead to uncontrolled proliferation — most commonly, the white tumouris visible through the pupil(leukocoria) or blocks vision.4When noticed early, prompt treatment usually cures. Delayed diagnosis can lead to incurableinvasion ofthe optic nerve and brain,or to metastasis which issometimescuredby extensiveintervention.

There are a number of challenges to related to our understanding of retinoblastoma and delineating its optimal clinical management. The cell of origin of retinoblastoma has eluded us thus far; it’s identity could unlock our understanding of why the infant retina is to uniquely predisposed to malignancy upon RB1 loss. The evidence base for clinical management is weak, yet rRigorous clinical trials in retinoblastoma are difficult for multiple reasons:too few patients in countries with capacity to conduct clinical trials; often complex presentation (two eyes withdiffering severity); too few patients to interest the pharmaceutical industry; multidisciplinary collaborationrequired but difficult, given often-opposing schools of thought in the field; and high societal value oneyes and vision imposes considerations beyond curing the cancer. New technologies showing dramatic primary treatment response have been quickly embraced, without rigorous clinical trials.

The Internet has opened many avenues: parents diagnose retinoblastoma themselves; colleagues share patients around the globe; centres of excellence aremapped (Figure2); and a common database for all children no matter where they live is within reach.These developmentsare set to empower a learning health system that build and evidence base for care.This retinoblastoma Primerintroducesunprecedented new science, ideas, therapies and global collaboration.

[H1] Epidemiology (HD,GC, JZ, FN)

[H2] Globalpatients, resources and outcomes

The expected number of patients with retinoblastoma annually per country can be calculated by multiplying theglobalretinoblastoma incidence (1 in 16,000-18,000 live births) by forecast births (Table 1).5-7This predicts approximately 8,000 new cases each year.

Of all affected children, 11%reside in high-income,69% in middle-incomeand 20% inlow-income countries. Although the prevalence is higher in middle- to low-income countries, most retinoblastoma treatment centres are inmiddle- and high-income countries,creatinga gap in healthcare access(Table2).Consistent with income being a surrogatefor non-economicmeasures of standard of living, retinoblastoma in low-income countries is associated with low patient survival(~30%8, 9) compared with high-income countries (>95%10), but comprehensive nation-widedata is lacking.

Poor outcome correlates with late presentation, difficulty accessing retinoblastoma-specific health care, and socio-economic issues leading to poor compliance, includingfamily refusal to remove the affected eye and abandonment of therapy.11-13Without timely diagnosis and appropriate treatment, difficult-to-cure metastatic disease may occur. Fortunately, with early diagnosis,14 many eyes can be safely treated to support a lifetime of good vision, pointing to key elements for global focus: awareness, collaboration, and affordable expert care.

[H2] Solutions for global retinoblastoma

A number of initiatives address the inequality in retinoblastoma treatment between developing and developed worlds.In 2009, the Canadian National Retinoblastoma Strategy (NRbS) published the first-ever retinoblastoma clinical practice guidelines,10 adapted by the Kenyan National Retinoblastoma Strategy (KNRbS) and published in 2014 in partnership with the Kenyan Ministry of Health.15The KNRbS achieved standardization of pathology processing and reportingto support treatment decisions and discussion of prognosis with families. Adoption of upfront (first-line)enucleation with implants and immediate prosthetic eyes (sourced from India), parent to parent interactions to allay uninformed fears, and standardization of information provided to parents has reduced the rate of non-compliance with treatment.15, 16In 2013, the Paediatric Oncology in Developing Countries Committee of the SIOP (International Society of Paediatric Oncology) published a consensus guideline for the management of retinoblastoma in countries with limited resources17 with clear ideas that can shape resource development. The Mexican National Strategy18and the Brazilian SOBOPE (Brazilian Society of Pediatric Oncology)19 guidelines are also applicable at a national level with governmental support for treatment.

Peer-to-peer collaborations and twinning programmeshave built a framework for knowledge and expertise exchange, filling gaps in specialized training, and source donations of equipment and resources with the ultimate aim of sustainable local capacity.20 Retinoblastoma-specific twinning programmes include partnerships between St. Jude’s Children’s Research Hospital (USA) and the Middle-East,21 Central America22 and Mexico,18and between the Institut Curie and a centre in Bamako, Mali.23The Central American Association of PediatricHematology Oncology (AHOPCA) created a cooperative group and implemented multicentre protocols for retinoblastoma treatment,22, 24, 25 a major achievement, not yet paralleled in developed countries. The AHOPCA funding is now 90% local. Twinning programmesbenefit from strong participation of both non-governmental and governmental organizations. However, where government may prove volatile and unpredictable,sustainable local capacity is a challenge. Less formal cooperation between developed and less developed countries can also result in highly efficient programmes.26, 27

A successful national strategy hasbeen developed in China.About 1,100 newly diagnosed cases are forecast annually, scattered over 32 provinces,imposing high travel costs. Before 2005, enucleation was the only available treatment for most children. For better treatment options and follow-up, centres classified by expertise and resources were established in 28 hospitals covering 25 provinces (over 90% of the population)28.The improved efficiency and collaborationfrom 2006 to 2014 has led to standardized classification and treatment of 2,097 newly diagnosed patients with retinoblastoma on commonprotocols with gains in survival (unpublished data, JZ).In Argentina, a single centre with high expertise coordinates affiliated retinoblastoma clinics. Thiscollaboration between governmental hospitals, local and international NGO’s with prospective protocols has significantly improved survival in two decades.29 In addition, translational research and population-based incidence and survival studies are reported.30, 31

In a rapidly changing world, Internet based strategies are facilitatingglobal collaboration. The One Retinoblastoma World map (Figure 2) leads families to the nearest centre with known expertise.The retinoblastoma-specific point-of-care database (eCancerCareRB, eCCRB) summarizes the medical record for retinoblastoma for each child. eCCRBon the Internet is freely accessible to every retinoblastoma centre with local ethics and privacy approvals. The Disease-specific, electronic Patient Illustrated Clinical Timeline (DePICT) is language-independent, well understood by parents, and supports fully informed care choices.32With guardian consent, fully identified data is accessible to those in the circle of care of each child.Ultimately, the de-identified data will be fuel for powering a global learning health system for retinoblastoma.

[H1] Mechanisms/pathophysiology (TWC, DC, HD, BLG)

The search for the genetic basis of retinoblastoma provided the two-hit model of tumor suppressor gene inactivation.In 1971, Knudson discovered that the age of diagnosis of retinoblastoma isconsistent with one rate-limiting event in bilaterally affected patients (who have heritable disease), and two events in unilaterally affected patients with no family history (usually non-heritable).33Othersproposed that heritable retinoblastomas result from a germline mutation (‘first hit’) and an acquired somatic mutation (‘second hit’), and non-heritable retinoblastomas arise when two somatic mutations in the same transformation suppressor gene in a susceptible cell(Figure 3).34Chromosomal deletions in some patients pointed to a chromosome 13q14 locus. Loss of heterozygosity of 13q14 polymorphic markers in 70% of retinoblastoma tumours35 implied that the second hit involved the same locus. The breakthrough came when one retinoblastoma was missing the sequence of a 13q DNA clone,36 that turned out to be a conserved, exonic sequence of RB1.36-39

RB1 is a large (190 kb) gene with 27 exons, encoding a 4.7 kb mRNA that translates into a 928 amino acid protein, pRB. Many modifications impair pRBfunction, including point mutations, promoter methylation, and small and large deletions.40 The A/B “pocket” region41harbours most missense mutations.[A1]

The pRB functions that normally suppress retinoblastoma within the retina remain unknown. pRB is best known as a cell cycle regulator that binds to E2F transcription factors to repress cell proliferation-related genes. Hyperphosphorylation of pRB by cyclin-dependent kinases in response to mitogenic signals normally relieves repression and promotes the G1 to S phase transition.pRB loss relieves this suppression in the absence of mitogenic signals to enable tumourigenesis. It is tempting to speculate that pRB is primarily needed to suppress E2F.41, 42Several“low penetrance” RB1 mutations encode proteins with minimal ability to bind E2F that result infewer tumours than RB1 null alleles.43, 44Suchdefective E2F-binding alleles may function to block retinoblastoma development through an E2F-independent mechanism. pRBalso up-regulates p27, implicated in cell differentiation, apoptosis, and genomic integrity.41IncreasedpRB expression is associated with decreasing p27 during cone precursor maturation.42pRB N-terminus functions may also be important.45, 46

BiallelicRB1 inactivation is necessary to initiate most retinoblastomas, but is not sufficient; the benign retinal lesion, retinoma, has lost both RB1-/-alleles(Figure 3).3 Further genetic or epigenetic changes are likely needed formalignant transformation.47 Comparative genomic hybridization studies identified several candidate retinoblastoma oncogenes(mitotic kinesin KIF14 and p53 regulator MDM4 (1q32), transcription factors E2F3 and DEK (6p22), and the onco-miR clusters miR-106b~25 (7q22.1) and miR-17~92 (13q31)) and the cadherin-11 (CDH11) tumour suppressor gene (16q22 loss). Whole-genome sequencing identified inactivating mutations in the transcriptional corepressorBCOR.48 Epigenetic alterationsmightdrive retinoblastoma formationby inducing H3K4me3 and H3K9/14ac marks and expression of the SYK oncogene.48Other genes and microRNAsalso show altered expression in retinoblastoma compared to normal retina.47 Gene expression profiles may segregate RB1-/-retinoblastomas either into two subtypes or into a spectrum of phenotypescorrelating with histologic and cytogenetic aberrations.49, 50

Nearly all retinoblastomas have mutation of both RB1 alleles, but 1.4% of unilateral tumours show evidence ofRB1 mutation, buthavehigh-level amplification of the oncogene MYCN (MYCNA).51These RB1+/+MYCNAtumours are diagnosed in much younger children than unilateral RB1-/- tumourswith distinct histology, reflecting a unique subtype.Another 1.5% of unilateral non-familial unexplained retinoblastomas have apparently normal both RB1and MYCNgenes.51

[H2] The retinoblastoma cell of origin

Since pRB is expressed in most if not all cells,the retina’s unique sensitivity to pRB loss is perplexing. Diverse retinal cell type markers and RNAs suggest a pluripotent cell of origin, aberrant expression of oncogenic transformation or normal intra-tumourRB1+/+cells.49, 52Mouse models have not clarified the cell-of-origin, since they require loss of Rb1 plus p107, p130, or p27,53express different retinal markers, and may not extrapolate to humans.54

Thetopographic distribution of emerging retinoblastomasmimics the horizontal visual streak characteristic of red and green cones, evidence for a cone precursor to be the cell of orgin (Figure 4).44RB1-/- retinoblastomas show consistently express cone photoreceptor but not other retinal cell type-specific proteins; and maturing cone precursors have unusually high expression ofoncoproteins (MDM2 and N-myc)that could collaborate with RB1 loss.52Moreover, experimental depletion of RB1 induced cone precursor cell proliferation in vitro and in vivoexperimental tumours typical of differentiated retinoblastomas.55 Proliferation depended upon high levels of N-Myc and MDM2, cone-specific transcription factors RXR and TR2, and down-regulation of p27associated with cone precursor maturation.42, 55pRBmay counter an oncogenic programme associated with cone precursor maturation.

Notably,smalltumours detected via optical coherence tomography (OCT) appear to be centred in the inner nuclear layer of the retina, not the outer nuclear layer where mature cones reside.56 However, they also extendinto the outer nuclear layer (Figure4c). Perhapsblood vessels and retinal astrocytes in inner retina promote retinoblastoma cell growth in the inner nuclear layer56 and in vitro.57 However, gene expression studies suggest several different classes of RB1-/- retinoblastomas that may arise in different cell types.50

[H2] Translating knowledge of pathogenesis

Understanding retinoblastoma molecular pathways could lead to treatment and prevention opportunities. Oncoproteins such as N-myccould be targeted58 in both RB1+/+MYCNAtumours51 and RB1-/- tumours that have progressed toN-myc-dependence.52 Furthermore, studies based on molecular discoveriesshowed thatexposing retinoblastoma-prone murine fetuses to small molecule inhibitors of E2f or Cdk inhibited subsequent tumorigenesis without disrupting normal retinal development.53

These new translational opportunities require cell lines and animal models that accurately reflect retinoblastoma cell responses. Most in vitro studies have used two old cell lines, but many othersare available.59 Primary xenografts in immunodeficient mice60are useful48but will not reflect the cell and immune environment of natural retinoblastoma, potentially limiting translation to patients.Genetically engineered mouse models can also be used to assess novel treatments.61 While Rb1+/- mice do not develop retinoblastoma, retinal deletion of Rb1 (using Pax6a, Nestin, or Chx10 promoters) in p107, p130 or p27mutant backgroundsachieves retinal tumour formation.54Such models have been used to examine genetic interactions in vivo, such as the role of microRNA miR-17~92 overexpression in Pax6a-Cre;Rb1lox/lox; p107-/- mice.62 However, experimentalmurine tumours have different collaborating mutations and cell type origin, reducing their ability to predict human treatment responses. Viral oncoproteins, such as Simian Virus 40 T-antigen,promote murine retinoblastoma specifically in developing Müller cells.63, 64There is lots of room for models that more precisely simulate retinoblastoma1 pathogenesis, to develop novel therapies that willtarget RB1-/- cancers.

[H1] Diagnosis, screening and prevention(HD, GC, JZ, BLG)

[H2] Clinical diagnosis

Diagnosis of retinoblastoma is usually clear from presenting signsand clinical examination.4The most common sign is leukocoria (white pupil)(Figures5).When parents report a strange reflection in the child’s eye, retinoblastoma should be at the top of the differential diagnosis. The second most common sign is strabismus (misaligned eyes) whencentral vision is lost. Advanced disease may present with iris colour change, enlarged cornea and eye from pressure, or non-infective orbital inflammation.Very late, the eye may bulge from the orbit, a common presentationwhere awareness and resources are inadequate.

Diagnosis of retinoblastoma does not rely on pathology, since biopsy incurs risk of metastasis.65With the pupil pharmacologically dilated and the indirect ophthalmoscope,the diagnosis is usually clear to an eye specialist.Calcification, characteristic of retinoblastoma, is detected by ultrasonography(b-scan). MRI is used to assess invasion the optic nerve and trilateral retinoblastoma (pinealoblastoma and primitive neuroectodermalintracranial tumours associated with RB1 mutations) (Figure 5). CT scansare nowavoided because radiation induces second primary cancers in people carrying RB1 mutations.66

Detailed retinal examination under general anaesthesiais required to distinguish the differential diagnoses (Coats disease, persistent fetal vasculature, and vitreous haemorrhage) and classifythe severity of intraocular disease.Accurate fundus drawing is essential to map tumor burden and location, and is some low-resources settings is the only available form of imaging. Where resources allow, tThe wide-angle, hand-held fundus camera is used to view and record all the retina, while as an expert depresses the sclera to bring the most anterior retina into view, while watching foroptic nerve arterial pulsations indicative of excessive pressure.Very useful arehigh frequency (50 MHz)ultrasound biomicroscopy,67and OCT68to discover invisible tumours in infants with familial disease.Good imaging and documentationof the whole retina supports eye classification and cancer staging, documents treatment responses, supports consultation with colleagues, and helps parents understand treatment options.