White eye it’s a sign of pathology


A team of researchers from Baylor University, with assistance from staff at the Massachusetts Eye and Ear Infirmary, Harvard Medical School and the Dana-Farber Cancer Institute has developed and tested a smartphone app that is able to detect “white eye” in children by analyzing stored photographs.

In their paper published in the journal Science Advances, the group describes how the app was developed and tested, and how well it works.

Most everyone has seen pictures of people seemingly possessed by the devil because their pupils glow red – this is caused by light bouncing off their retinas.

However, such pictures sometimes produce white instead of red retinas.

Sometimes it can happen due to ambient lighting conditions, but other times, it can indicate an eye ailment. Such problems can include retinoblastoma, a type of eye cancer, retinopathy, or even cataracts.

The idea for an app that could detect white eye came from the experience of one of the researchers, Brian Shaw, and his son, who developed retinoblastoma and subsequently lost an eye.

The team developed the app and made it available to the public back in 2014, but it was not until more recently that the team decided to test the app to see how well it works.

The researchers started by looking at thousands of photographs of Shaw’s son from birth until diagnosis.

They found the app was able to spot white eye in the child when he was just 12 days old – early enough that the eye might have been saved had it been diagnosed.

The team then tested the app on 53,000 images of 40 children stored on volunteer’s smartphones. Half of those in the test group had children with already diagnosed eye ailments.

The researchers report that the app found 16 of the 20 children with an eye disease. Perhaps even more importantly, it detected problems in the eyes of the children when they were younger—more than a year before diagnosis.

They note also that the app (called White Eye Detector) is free, allowing parents to download and use it whenever they wish.

App can detect “white eye” in children’s photos to spot possible problems
The “white eye detector” app scans photographs for symptoms of childhood eye disorders. Credit: Munson et al., Sci. Adv. 2019

Retinoblastoma is the most common intraocular malignancy of childhood and infancy accounting for 3% of all pediatrics cancers. It is caused by inactivation of RB1 genes commonly known as tumor suppressor gene[1].

Incidence of retinoblastoma ranges approximately worldwide at one case per 15 000-20 000 live births, which corresponds to about 9000 new cases every year[2]. There is no racial or gender predisposition.

The sign of retinoblastoma is a white pupil, called leukocoria, strabismus, painful blind eye and loss of vision[3]. With the advancement in treatment and multidisciplinary approach, eye salvage is possible in group A through D intraocular tumours[4]. However, in group E tumours and group D (unilateral cases), enucleation always remains the choice of treatment. Higher incidences of histopathologic risk factors are reported from enucleated eyes from developing countries as opposed to developed countries[5].


Knudson’s putative tumor suppressor gene was cloned by Weinberg’s lab by studying genetic lesions and chromosomal aberrations in families with a history of retinoblastoma. This gene was the first human tumor suppressor gene to be cloned and it was named as RB1[6]. Initially, researchers considered that RB1 was important only for retinoblastoma susceptibility but Harbour et al[7] found that the RB1 gene was also mutated in lung cancer.

India has the largest number of retinoblastoma cases with an estimated 1500 new cases annually[8]. It was the first tumour in which cancer genetics was revealed[9]. Cases with heritable retinoblastoma (48%) carry a germline mutation in the RB1 gene and are likely to develop secondary cancers later in life like bone and soft tissue sarcomas, melanoma, brain tumours, etc. 

They have a 50% risk of transmitting their germline RB1 mutation to their offspring[10]. Cases with nonheritable retinoblastoma have normal RB1 gene. Some of the retinoblastoma cases are caused by RB1 gene mutation while others are caused by somatic amplification of the MYCN gene[11]. Recently, genetic laboratories have found that retinoblastoma may arise when the MYCN oncogene is amplified even in the presence of non-mutated RB1 genes. These cases are relatively rare, occuring in <3% of unilateral retinoblastoma cases[12].

Only 6% are familial while 94% are sporadic in newly diagnosed cases of retinoblastoma. All the cases of bilateral retinoblastoma involve germinal mutations[13]. Almost 15% of unilateral sporadic retinoblastoma is caused by germinal mutations affecting only one eye while the 85% are sporadic.

Knudson proposed the two hit hypothesis in 1971[9]. Knudson stated that two chromosomal mutations are needed for developing retinoblastoma. The initial hit is a germinal mutation, which is inherited and is found in all cells in hereditary retinoblastoma. The second hit grows in the somatic retinal cells leading to the development of retinoblastoma.

Therefore, hereditary cases are subjected to the development of non-ocular tumors such as osteosarcoma. In unilateral sporadic retinoblastoma, both the hits occur during the development of the retina and are somatic mutations[14]. Therefore, there is no risk of second nonocular tumors.

Genetic testing for RB1 mutations and counselling of the patient can improve disease outcome and management. There are definitive molecular tests which help in identifying children and their relatives who are at high risk for retinoblastoma, and need to be followed closely for the disease[15]. Recently, combinational approach of multiplex ligation-dependent probe amplification assay, deletion screening, direct sequencing, copy number gene dosage analysis and methylation assays provides mutational spectrum of RB1 gene mutation in retinoblastoma patients[16].



The gross appearance of retinoblastoma at cut section of the eye is somewhat variable, reflecting the stage of the disease at enucleation. The tumor has a white, encephaloid or brain-like appearance, with chalky areas of calcification and yellow necrotic areas[17]. The presence of calcium is often accentuated in eyes that have had prior radiotherapy or chemotherapy. Gross examination of the eyeball in the laboratory involves a total of 4 blocks. One block is the eyeball section with the optic nerve. Two blocks should contain the calottes. The fourth block consists of the resected margin of the optic nerve.


On microscopic examination, retinoblastoma reveals a tumor composed of small hyperchromatic cells with a high nuclear to cytoplasmic ratio with large areas of necrosis and multifocal area of calcifications. Tumour differentiation are categorized into well differentiated [>50% known as Homer-Wright (HW) rosettes] or poorly differentiated [<50% known as Flexner-Wintersteiner (FW) rosettes][18].

In 2014, new rosettes were found which were comparatively larger than FW and HW and has an unusal anterior segment involvement[19]. Necrosis in the tumor is graded as none (<25%), mild (25%-50%), or extensive (>50%). Optic nerve invasion is graded as prelaminar, postlaminar and invasion of the resected margin. Postlaminar invasion is defined as tumour invasion beyond the lamina cribrosa of optic nerve[20].

Choroidal involvement is divided into focal invasion defined as a tumour focus of less than 3 mm in any diameter (thickness or width) or massive invasion defined as invasive focus of tumour measuring 3 mm or more in any diameter as per the Retinoblastoma Staging Working Group[21].

Artifactual seeding of tumour cells is seen at times in the sections which pose a problem to the pathologist. These are composed of small groups of tumour cells usually with many necrotic cells present within natural spaces of the eye. In contrast true tumour invasion comprises of solid nests of tumour cells with pushing or infiltrating borders without necrosis.

Histopathological Prognostic High Risk Factors

The survival and management of high risk retinoblastoma has improved by identification of high-risk factors and appropriate adjuvant therapy. Histopathological high risk factors (HRFs) are evaluated and identified after enucleation for predicting metastasis. Prognostic factors like massive choroidal invasion, retrolaminar invasion and involvement of resected end of optic nerve, iris and ciliary body involvement, anterior chamber involvement, scleral and extrascleral involvement by tumour cells are associated with a greater risk of orbital recurrence and predictive of metastasis. There is a still debate regarding anterior chamber as a high-risk factor for retinoblastoma. Recently, Sreelakshmi et al[22] concluded in their study that anterior chamber seeds do not, by themselves, constitute an independent risk factor for metastasis in retinoblastoma.

The reported incidences of HRFs are 7% to 56% for invasion of retrolaminar optic nerve and optic nerve to the transaction line; 12% to 42% for choroidal involvement; and 3% to 30% for scleral and extrascleral spread. Kashyap et al[23] described variousclinical features like older age at presentation, longer lag period, presence of hyphema, pseudohypopyon, staphyloma, and orbital cellulitis.

These factors were associated with occurrence of HRFs and may be a useful indicator for considering adjuvant chemotherapy especially in developing countries. Also poorly differentiated retinoblastomas present at a later age and are associated with presence of multiple HRFs and necrosis[18].

Cases with presence of HRFs need systemic adjuvant chemotherapy which improves the survival of children at risk for metastatic disease[24]. Therefore, histopathologic HRFs can provide important basis for clinicians to determine treatment plan.

Pathological Tumor, Node, Metastasis Classification

Tumor, node, metastasis (TNM) classification is developed by the American Joint Commission on Cancer (AJCC) and the Union International Control Cancer (Table 1)[25]. Retinoblastoma is the first cancer in which role of germline predisposition is recognised by incorporating stage category “H” into the AJCC classication[26]Table 2 describes the AJCC 2017 8th edition tumor, node, metastasis, heritable trait (TNMH) clinical (c) and pathological (p) staging system which is known to be the first evidence-based system for predicting overall prognosis of both eye(s) and patients[27][28].

Table 1

AJCC pathological classification (pTNM)

Primary tumor (pT)
pTXPrimary tumor cannot be assessed
pTONo evidence of primary tumor
pT1Tumor confined to the eye with no optic nerve or choroidal invasion
pT2Tumor with minimal optic nerve or choroidal invasion
pT2aTumor superficially invades optic nerve head, or tumor exhibits focal choroidal invasion but does not extend past lamina cribrosa
pT2bTumor superficially invades optic nerve head and tumor exhibits focal choroidal invasion but does not extend past lamina cribrosa
pT3Tumor with significant optic nerve and/or choroidal invasion
pT3aTumor invades optic nerve past lamina cribrosa but not to surgical resection line, or tumor exhibits massive choroidal invasion
pT3bTumor invades optic nerve past lamina cribrosa but not to surgical resection line and exhibits massive choroidal invasion
pT4Tumor invades optic nerve to surgical resection line or exhibits extra-ocular extension elsewhere
pT4aTumor invades optic nerve to resection line, but no extra-ocular extension identified
pT4bTumor invades optic nerve to resection line, and extra-ocular extension identified
Regional lymph nodes (pN)
pNXRegional lymph nodes cannot be assessed
pNONo regional lymph node metastasis
pN1Regional lymph node involvement (preauricular, cervical)
pN2Distant lymph node involvement
Metastasis (pM)
pMXPresence of metastasis cannot be assessed
pM0No distant metastasis
pM1Metastasis to sites other than central nervous system
pM1aSingle lesion
pM1bMultiple lesions
pM1cCentral nervous system metastasis
pM1dDiscrete masses without leptomeningeal and/or cerebrospinal fluid involvement
pM1eLeptomeningeal and/or cerebrospinal fluid involvement

Table 2

AJCC cTNMH retinoblastoma staging

Primary tumor (cT)
cTXUnknown evidence of intraocular tumour
cTONo evidence of intraocular tumor
cT1Intraocular tumour(s) with subretinal fluid ≤5 mm from the base of any tumour
cT1aTumours ≤3 mm and further than 1.5 mm from the disc and fovea
cT1bTumours >3 mm or closer than 1.5 mm to the disc and fovea
cT2Intraocular tumour(s) with retinal detachment, vitreous seeding or subretinal seeding
cT2aSubretinal fluid >5 mm from the base of any tumour
cT2bTumours with vitreous seeding and/or subretinal seeding
cT3Advanced intraocular tumour(s)
cT3aPhthisis or pre-phthisis bulbi
cT3bTumour invasion of the pars plana, ciliary body, lens, zonules, iris or anterior chamber
cT3cRaised intraocular pressure with neovascularization and/or buphthalmos
cT3dHyphema and/or massive vitreous haemorrhage
cT3eAseptic orbital cellulitis
cT4Extraocular tumour(s) involving the orbit, including the optic nerve
cT4aRadiological evidence of retrobulbar optic nerve involvement or thickening of the optic nerve or involvement of the orbital tissues
cT4bExtraocular tumour clinically evident with proptosis and orbital mass
Regional lymph nodes (cN)
cNXRegional lymph nodes cannot be assessed
cN0No regional lymph nodes involvement
cN1Evidence of preauricular, submandibular, and cervical lymph node involvement
Distant metastasis (M)
cM0No signs or symptoms of intracranial or distant metastasis
cM1Distant metastasis without microscopic confirmation
cM1aTumour(s) involving any distant site (e.g. bone marrow, liver) on clinical or radiological tests
cM1bTumour involving the central nervous system on radiological imaging (not including trilateral retinoblastoma)
pM1Distant metastasis with microscopic confirmation
pM1aHistopathological confirmation of tumour at any distant site (e.g. bone marrow, liver, or other)
pM1bHistopathological confirmation of tumour in the cerebrospinal fluid or central nervous system parenchyma
Heritable trait (H)
HXUnknown or insufficient evidence of a constitutional RB1 gene mutation
H0Normal RB1 alleles in blood tested with demonstrated high sensitivity assays
H1Bilateral retinoblastoma, retinoblastoma with an intracranial central nervous system midline embryonic tumour (e.g. trilateral retinoblastoma), patient with family history of retinoblastoma, or molecular definition of constitutional RB1 gene mutation


Ongoing studies via higher resolution genomic technologies, gene expression profiling, direct gene sequencing, multiplex-polymerase chain reaction, mi-RNA microarray profiling, next-generation sequencing (NGS), microsatellite analysis for loss of heterogeneity, and in-situ hybridization for chromosomal aberrations will continue to facilitate our exploration into the molecular intricacies of retinoblastoma and results in newer therapeutic approaches.

The discovery of proto-oncogenes transformed our insight into mechanisms of cancer. More recent studies shows that retinoblastoma tumors may differ in the mutagenic pathway as some of retinoblastoma tumors are caused by RB1 mutation[29][32] while some can also be initiated by amplification of MYCN proto-oncogene.

Singh et al[33] demonstrated prognostic significance of CDC25 phosphatases and polo-like kinases in retinoblastoma. They suggested that expression of CDC25B might be used as a potential prognostic marker in the pathogenesis of retinoblastoma and contribute to the development of the disease by causing genomic instability through deregulation of cell division. In their study, PLK1 was more frequently expressed and deregulated in poorly differentiated retinoblastoma tissue as compared to PLK3 protein that might serve as a poor prognostic marker in retinoblastoma[20].

Evasion of apoptosis is a hallmark of human cancers that leads to cancer development, progression and treatment resistance. The Bcl-2 family members are important regulators of the mitochondrial pathway of apoptosis. Bax and Bcl-2 are proteins that regulate programmed cell death and apoptosis. Recently, Singh et al[34] revealed higher expression of Bcl-2 in 66% of cases whereas Bax expression was found only in fewer cases (30%) of retinoblastoma tissue by immunohistochemistry, mRNA and Western blotting techniques. According to the author, differential expression of apoptotic regulatory proteins might represent poor response to patient outcome and have potential for tumor invasiveness.

Grotta et al[35] used a combined approach of next-generation sequencing (NGS) and RB1 custom array-comparative genomic hybridization (aCGH) on a cohort of retinoblastoma patients. NGS and RB1 custom aCGH have demonstrated to be an effective combined approach in order to optimize the overall diagnostic procedures of retinoblastoma. Devarajan et al[36] demonstrated for the first time that targeted next generation sequencing is an efficient approach for the identification of wide spectrum of pathogenic variants in retinoblastoma patients. Using this approach, an array of pathogenic variants including single nucleotide variants, InDels (small insertions/deletions) and copy number variations were detected in retinoblastoma patients. This comprehensive approach reduces the time and number of assays required for the detection of pathogenic variants by conventional methods which is sensitive (0.97) and efficient for RB1 screening.

The application of genomics to the study of cancer is rapidly shifting toward the analysis of tissue samples to discover new biomarkers for early detection of cancers. Mitochondria have been implicated in tumor progression, cell differentiation, and apoptotic pathways. The identification of mitochondrial DNA mutations and its associated proteins as a biomarker has been used to help understand not only gene function but also the underlying molecular mechanisms of mitochondrial biology in retinoblastoma. This strategy relies on the hypothesis that if mutations in mtDNA cause physiological aberrations specifically in a particular tissue, the gene is more likely to be selectively expressed in that tissue. Currently, the role of mitochondria in retinoblastoma biology is still poorly understood.

Recently, Singh et al[37] have described and analyzed the morphological changes of mitochondria in retinoblastoma tumor by transmission electron microscopy. Poorly differentiated retinoblastoma cases showed fewer mitochondria, scant cytoplasm, disorganized organelles (mitochondria), and necrosis, whereas well-differentiated retinoblastomas had larger number of mitochondria and more organized organelles. Understanding the structural and functional characteristics of mitochondria in retinoblastoma might be essential for the design of future therapeutic strategies. They have also studied the expression of mitochondrial oxidative phosphorylation complexes in retinoblastoma tumor tissues. Among all the complexes, loss of mitochondrial complex I immunoexpression proved to be a useful independent prognostic biomarker to identify high-risk retinoblastoma patients[38]. Role of mitochondrial DNA and its protein biomarkers necessitate careful experimentation to adequately assess its contribution in retinoblastoma which might prove of diagnostic and prognostic value, and serve as a basis for the development of better long term therapeutic strategies. This provides an insight into molecular mechanisms of mitochondrial dysfunction, and also helps to find novel cancer biomarkers in retinoblastoma.

Sangeetha et al[39] investigated lipogenesis-dependent survival of retinoblastoma cancer cells and the associated molecular pathways in fatty acid synthase (FASN) silenced retinoblastoma cells and revealed that FASN silencing reduced the invading property of retinoblastoma cancer cells by scratch assay. Venkatesan et al[40] studied computational and in vitro investigation of miRNAs-gene regulation in retinoblastoma pathogenesis by an in silico approach. They concluded downregulation of miR-486-3p and miR-532-5p in primary retinoblastoma tissues, which might implicates their role in tumorigenesis.

Cytoplasmic expression of FOXO3a (transcription factor) has been found to be associated in pathogenesis of retinoblastoma. Relocation of FOXO3a from cytoplasmic to nucleus activates non-mutated retinoblastoma and might be a therapeutic target for retinoblastoma[41]. Reactive oxygen species and free radicals are associated with cancer development and its progression, which might suggest potential avenues of therapeutic intervention. Expression of NOX4 protein might be a source of reactive-oxygen species production in tumor cells, leading to oxidative stress and associated with less overall survival rate in retinoblastoma[42].

Sirt1 (Sirtuin1) is the most important protein among all the sirtuins as it involves multiple factors that are highly relevant to cancer. Batra et al[43] found high expression of Sirt1 in retinoblastoma tumors but it was not associated with any high-risk histopathological factors. Similarly, Sirt2 and Sirt6 were also expressed in tumor cells along with various normal structures of the remaining ocular tissues[44]. Apart from the expression studies, mass spectrometry-based quantitative proteomic approach was also implemented to identify differentially expressed proteins in which mitochondrial dysfunction and lipid metabolism pathways found to be deregulated in retinoblastoma tumor[45].

Recently, an elevated expression of PDK1 protein levels has been validated in retinoblastoma tumors especially in vitreous seeds and hypoxic regions. Inhibition of PDK1 gene in retinoblastoma cell lines demonstrated reduced cell growth and increased apoptosis, which might be a potential future therapeutic target in retinoblastoma[46].

Journal information:Science Advances


Please enter your comment!
Please enter your name here

Questo sito usa Akismet per ridurre lo spam. Scopri come i tuoi dati vengono elaborati.