Prediction of molecular subclasses of uveal melanoma by deep learning using routine haematoxylin-eosin-stained tissue slides.

AIMS: Uveal melanoma has a high propensity to metastasize. Prognosis is associated with specific driver mutations and copy number variations, and these can only be obtained after genetic testing. In this study we evaluated the efficacy of patient outcome prediction using deep learning on haematoxylin and eosin (HE)-stained primary uveal melanoma slides in comparison to molecular testing. METHODS: In this retrospective study of patients with uveal melanoma, 113 patients from the Erasmus Medical Centre who underwent enucleation had tumour tissue analysed for molecular classification between 1993 and 2020. Routine HE-stained slides were scanned to obtain whole-slide images (WSI). After annotation of regions of interest, tiles of 1024 × 1024 pixels were extracted at a magnification of 40×. An ablation study to select the best-performing deep-learning model was carried out using three state-of-the-art deep-learning models (EfficientNet, Vision Transformer, and Swin Transformer). RESULTS: Deep-learning models were subjected to a training cohort (n = 40), followed by a validation cohort (n = 20), and finally underwent a test cohort (n = 48). A k-fold cross-validation (k = 3) of validation and test cohorts (n = 113 of three classes: BAP1, SF3B1, EIF1AX) demonstrated Swin Transformer as the best-performing deep-learning model to predict molecular subclasses based on HE stains. The model achieved an accuracy of 0.83 ± 0.09 on the validation cohort and 0.75 ± 0.04 on the test cohort. Within the subclasses, this model correctly predicted 70% BAP1-mutated, 61% SF3B1-mutated and 80% EIF1AX-mutated UM in the test set. CONCLUSIONS: This study showcases the potential of the deep-learning methodology for predicting molecular subclasses in a multiclass manner using HE-stained WSI. This development holds promise for advanced prognostication of UM patients without the need of molecular or immunohistochemical testing. Additionally, this study suggests there are distinct histopathological features per subclass; mainly utilizing epithelioid cellular morphology for BAP1-classification, but an unknown feature distinguishes EIF1AX and SF3B1.

ATRX Loss in the Development and Prognosis of Conjunctival Melanoma.

Metastatic disease is linked to TERT promoter mutations in conjunctival melanomas (CM). Both TERT promoter and ATRX mutations are associated with faulty telomere maintenance. This study aimed to determine the prognostic value of ATRX loss in conjunctival melanocytic lesions. Eighty-six conjunctival melanocytic lesions from the Rotterdam Ocular Melanoma Study group were collected. ATRX status and TERT promoter status were determined using immunohistochemical staining and molecular diagnostics, respectively. None of the nevi (n = 16) and primary acquired melanosis (PAM) without atypia (n = 6) showed ATRX loss. ATRX loss was found in 2/5 PAM with atypia without CM and in 8/59 CM. No cases with a TERT promoter mutation (n = 26) showed ATRX loss. Eight/eleven metastatic CM harbored a TERT promoter mutation, two other metastatic CM showed ATRX loss and one metastatic case showed no TERT promoter/ATRX alterations. In conclusion ATRX loss and TERT promoter mutations are only found in (pre)malignant conjunctival melanocytic lesions, with most metastatic cases harboring one of these alterations, suggesting that both alterations are associated with adverse behavior. Similar to TERT promoter mutations, ATRX loss may be used as a diagnostic tool in determining whether a conjunctival melanocytic lesion is prone to having an adverse course.

Genetic Background of Iris Melanomas and Iris Melanocytic Tumors of Uncertain Malignant Potential.

PURPOSE: Uveal melanoma (UM) is the most common primary intraocular malignancy in adults. Iris melanoma comprises 4% to 10% of all UMs and has a lower mortality rate. The genetic changes in iris melanoma are not as well characterized as ciliary body or choroidal melanoma. The aim of this study was to gain more insight into the genetic background of iris melanoma and iris nevi. DESIGN: Multicenter, retrospective case series. PARTICIPANTS: Patients diagnosed with iris melanoma or iris nevi who underwent surgical intervention as primary or secondary treatment. METHODS: Next-generation sequencing of GNAQ, GNA11, EIF1AX, SF3B1, BAP1, NRAS, BRAF, PTEN, c-Kit, TP53, and TERT was performed on 30 iris melanomas and 7 iris nevi. Copy number status was detected using single nucleotide polymorphisms (SNPs) included in the next-generation sequencing (NGS) panel, SNP array, or fluorescent in situ hybridization. BAP1 immunohistochemistry was performed on all samples. MAIN OUTCOME MEASURES: Mutation and copy number status were analyzed. Results of BAP1 immunohistochemistry were used for survival analysis. RESULTS: In 26 of the 30 iris melanoma and all iris nevi, at least 1 mutation was identified. Multiple mutations were detected in 23 iris melanoma and 5 nevi, as well as mutations in GNAQ and GNA11. Furthermore, 13 of 30 BAP1, 5 of 30 EIF1AX, and 2 of 30 SF3B1 mutations were identified in iris melanoma. No correlation between BAP1 status and disease-free survival was found. The iris nevi showed 1 EIF1AX and 3 BAP1 mutations. Two of the nevi, with a BAP1 mutation, were histologically borderline malignant. Mutations in NRAS, BRAF, PTEN, c-KIT, and TP53 were detected in 6 iris melanomas and 4 iris nevi. CONCLUSIONS: Mutations that are often found in uveal and cutaneous melanoma were identified in this cohort of iris melanomas and iris nevi. Therefore, iris melanomas harbor a molecular profile comparable to both choroidal melanoma and cutaneous melanoma. These findings may offer adjuvant targeted therapies for iris melanoma. There was no prognostic significance of BAP1 expression as seen in choroidal melanoma. Consequently, iris melanoma is a distinct molecular subgroup of UM. Histologic borderline malignant iris nevi can harbor BAP1 mutations and may be designated iris melanocytic tumors of uncertain malignant potential.

Uveal Melanoma Zebrafish Xenograft Models Illustrate the Mutation Status-Dependent Effect of Compound Synergism or Antagonism.

Purpose: Uveal melanoma (UM) is the most common primary intraocular malignancy with a high probability of metastatic disease. Although excellent treatment options for primary UM are available, therapy for metastatic disease remain limited. Drug discovery studies using mouse models have thus far failed to provide therapeutic solutions, highlighting the need for novel models. Here, we optimize zebrafish xenografts as a potential model for drug discovery by showcasing the behavior of multiple cell lines and novel findings on mutation-dependent compound synergism/antagonism using Z-Tada; an algorithm to objectively characterize output measurements. Methods: Prognostic relevant primary (N = 4) and metastatic UM (N = 1) cell lines or healthy melanocytes (N = 2) were inoculated at three distinct inoculation sites. Standardized quantifications independent of inoculation site were obtained using Z-Tada; an algorithm to measure tumor burden and the number, size, and distance of disseminated tumor cells. Sequentially, we utilized this model to validate combinatorial synergism or antagonism seen in vitro. Results: Detailed analysis of 691 zebrafish xenografts demonstrated perivitelline space inoculation provided robust data with high probability of cell dissemination. Cell lines with more invasive behavior (SF3B1mut and BAP1mut) behaved most aggressive in this model. Combinatorial drug treatment illustrated synergism or antagonism is mutation-dependent, which were confirmed in vivo. Combinatorial treatment differed per xenograft-model, as it either inhibited overall tumor burden or cell dissemination. Conclusions: Perivitelline space inoculation provides robust zebrafish xenografts with the ability for high-throughput drug screening and robust data acquisition using Z-Tada. This model demonstrates that drug discovery for uveal melanoma must take mutational subclasses into account, especially in combinatorial treatment discoveries.

Uveal melanoma modeling in mice and zebrafish.

Despite extensive research and refined therapeutic options, the survival for metastasized uveal melanoma (UM) patients has not improved significantly. UM, a malignant tumor originating from melanocytes in the uveal tract, can be asymptomatic and small tumors may be detected only during routine ophthalmic exams; making early detection and treatment difficult. UM is the result of a number of characteristic somatic alterations which are associated with prognosis. Although UM morphology and biology have been extensively studied, there are significant gaps in our understanding of the early stages of UM tumor evolution and effective treatment to prevent metastatic disease remain elusive. A better understanding of the mechanisms that enable UM cells to thrive and successfully metastasize is crucial to improve treatment efficacy and survival rates. For more than forty years, animal models have been used to investigate the biology of UM. This has led to a number of essential mechanisms and pathways involved in UM aetiology. These models have also been used to evaluate the effectiveness of various drugs and treatment protocols. Here, we provide an overview of the molecular mechanisms and pharmacological studies using mouse and zebrafish UM models. Finally, we highlight promising therapeutics and discuss future considerations using UM models such as optimal inoculation sites, use of BAP1mut-cell lines and the rise of zebrafish models.

FOXD1 Is a Transcription Factor Important for Uveal Melanocyte Development and Associated with High-Risk Uveal Melanoma.

Uveal melanoma (UM) is a deadly ocular malignancy, originating from uveal melanocytes. Although much is known regarding prognostication in UM, the exact mechanism of metastasis is mostly unknown. Metastatic tumor cells are known to express a more stem-like RNA profile which is seen often in cell-specific embryonic development to induce tumor progression. Here, we identified novel transcription regulators by reanalyzing publicly available single cell RNA sequencing experiments. We identified five transcription regulators of interest: ELL2, KDM5B, REXO4, RBFOX2 and FOXD1. Our most significant finding is FOXD1, as this gene is nearly exclusively expressed in high-risk UM and its expression is associated with a poor prognosis. Even within the BAP1-mutated UM, the expression of FOXD1 is correlated with poor survival. FOXD1 is a novel factor which could potentially be involved in the metastatic capacity of high-risk UM. Elucidating the function of FOXD1 in UM could provide insight into the malignant transformation of uveal melanocytes, especially in high-risk UM.

Case Report and Review of the Literature: Congenital Diaphragmatic Hernia and Craniosynostosis, a Coincidence or Common Cause?

Congenital diaphragmatic hernia (CDH) is a life-threatening birth defect that presents as either an isolated diaphragm defect or as part of a complex disorder with a wide array of anomalies (complex CDH). Some patients with complex CDH display distinct craniofacial anomalies such as craniofrontonasal dysplasia or craniosynostosis, defined by the premature closure of cranial sutures. Using clinical whole exome sequencing (WES), we found a BCL11B missense variant in a patient with a left-sided congenital diaphragmatic hernia as well as sagittal suture craniosynostosis. We applied targeted sequencing of BCL11B in patients with craniosynostosis or with a combination of craniosynostosis and CDH. This resulted in three additional BCL11B missense mutations in patients with craniosynostosis. The phenotype of the patient with both CDH as well as craniosynostosis was similar to the phenotype of previously reported patients with BCL11B missense mutations. Although these findings imply that both craniosynostosis as well as CDH may be associated with BCL11B mutations, further studies are required to establish whether BCL11B variants are causative mutations for both conditions or if our finding was coincidental.

Transient Expression of Lymphatic Markers in Retrobulbar Intraconal Orbital Vasculature During Fetal Development.

Purpose: The aim of this study is to investigate the presence of orbital lymphatic vessels during fetal and neonatal development and in adults using a panel of lymphatic markers. Methods: This was a retrospective observational case series. For analyzing lymphatic vessels, we used formalin-fixed paraffin-embedded enucleated eyes from 25 human fetuses between 13 and 24 weeks of gestation and postnatal eyes from 15 children and 5 adults. Immunohistochemical analysis of lymphatic vessels was performed for the markers: lymphatic vessel endothelial hyaluronic acid receptor-1 (LYVE-1), podoplanin (D2-40), Prospero-related homeobox gene-1 (Prox-1), pan-endothelial marker CD31, and blood vessel endothelium specific CD34. Results: Vasculature showing endothelial expression of LYVE-1, D2-40, Prox-1, and CD31 in combination with absence or weak expression of CD34, as would be expected for lymphatic vessels, was seen in 11 of 25 fetuses in an age range from 14 weeks to 23 weeks of gestation (44%). This lymphatic vascular staining pattern was also observed in 4 of 15 liveborn children (27%), all within 1 month of age, of which two were born prematurely at 32 and 34 weeks of gestation. Interestingly, an incomplete lymphatic staining pattern was observed in another 4 fetuses and two liveborn children of 4 months and 7 years old. No expression of lymphatic markers was observed in adult orbital vasculature. Conclusions: No retrobulbar intraorbital lymphatic vessels were observed in adults, however, we did observe transient expression of lymphatic markers in retrobulbar intraconal orbital vasculature during fetal and early neonatal development. The orbit may, therefore, be proposed to possess a full range of lymphatic plasticity.

Absence of Intraocular Lymphatic Vessels in Uveal Melanomas with Extrascleral Growth.

The aim of this study was to investigate the presence of intraocular lymphatic vessels in patients with uveal melanomas and extrascleral extension using a panel of lymphatic markers. The following immunohistochemical markers were analyzed: lymphatic vessel endothelial hyaluronic acid receptor-1 (LYVE-1), podoplanin (D2-40), prospero-related homeobox gene-1 (Prox-1), pan-endothelial marker cluster of differentiation 31 (CD31), and blood vessel endothelium-specific CD34. Lymphatic vessels were defined as a combination of staining of the following positive markers: LYVE-1, D2-40, Prox-1, and CD31; and no staining of the negative marker CD34. In total, 456 patients were enucleated; 16 of the 46 uveal melanomas with extrascleral extension were contained in stored paraffin tissue. Two samples of the 16 uveal melanomas showed focal positive intraocular vascular staining for LYVE-1 and co-expression of CD31 and CD34. Due to the lack of Prox-1 and D2-40, and positive expression of CD34, these cannot be classified as lymphatic vessels. In one case recruitment of an extraocular, intratumoral lymphatic vascular structure was observed in the periphery of the subconjunctival extrascleral extension. Intraocular lymphatic vessels are absent in uveal melanomas with extrascleral extension; however, we provide proof for recruitment of intratumoral lymphatics by uveal melanomas with extraocular extension from subconjunctival lymphatics that may explain the rare cases of regional lymphatic spread. A panel of antibodies is necessary to detect lymphatic vessels with high specificity.