Predicting Choroidal Nevus Transformation to Melanoma Using Machine Learning.

Purpose: To develop and validate machine learning (ML) models to predict choroidal nevus transformation to melanoma based on multimodal imaging at initial presentation. Design: Retrospective multicenter study. Participants: Patients diagnosed with choroidal nevus on the Ocular Oncology Service at Wills Eye Hospital (2007-2017) or Mayo Clinic Rochester (2015-2023). Methods: Multimodal imaging was obtained, including fundus photography, fundus autofluorescence, spectral domain OCT, and B-scan ultrasonography. Machine learning models were created (XGBoost, LGBM, Random Forest, Extra Tree) and optimized for area under receiver operating characteristic curve (AUROC). The Wills Eye Hospital cohort was used for training and testing (80% training-20% testing) with fivefold cross validation. The Mayo Clinic cohort provided external validation. Model performance was characterized by AUROC and area under precision-recall curve (AUPRC). Models were interrogated using SHapley Additive exPlanations (SHAP) to identify the features most predictive of conversion from nevus to melanoma. Differences in AUROC and AUPRC between models were tested using 10 000 bootstrap samples with replacement and results. Main Outcome Measures: Area under receiver operating curve and AUPRC for each ML model. Results: There were 2870 nevi included in the study, with conversion to melanoma confirmed in 128 cases. Simple AI Nevus Transformation System (SAINTS; XGBoost) was the top-performing model in the test cohort [pooled AUROC 0.864 (95% confidence interval (CI): 0.864-0.865), pooled AUPRC 0.244 (95% CI: 0.243-0.246)] and in the external validation cohort [pooled AUROC 0.931 (95% CI: 0.930-0.931), pooled AUPRC 0.533 (95% CI: 0.531-0.535)]. Other models also had good discriminative performance: LGBM (test set pooled AUROC 0.831, validation set pooled AUROC 0.815), Random Forest (test set pooled AUROC 0.812, validation set pooled AUROC 0.866), and Extra Tree (test set pooled AUROC 0.826, validation set pooled AUROC 0.915). A model including only nevi with at least 5 years of follow-up demonstrated the best performance in AUPRC (test: pooled 0.592 (95% CI: 0.590-0.594); validation: pooled 0.656 [95% CI: 0.655-0.657]). The top 5 features in SAINTS by SHAP values were: tumor thickness, largest tumor basal diameter, tumor shape, distance to optic nerve, and subretinal fluid extent. Conclusions: We demonstrate accuracy and generalizability of a ML model for predicting choroidal nevus transformation to melanoma based on multimodal imaging. Financial Disclosures: Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

Mitochondrial DNA variation and cancer.

Variation in the mitochondrial DNA (mtDNA) sequence is common in certain tumours. Two classes of cancer mtDNA variants can be identified: de novo mutations that act as 'inducers' of carcinogenesis and functional variants that act as 'adaptors', permitting cancer cells to thrive in different environments. These mtDNA variants have three origins: inherited variants, which run in families, somatic mutations arising within each cell or individual, and variants that are also associated with ancient mtDNA lineages (haplogroups) and are thought to permit adaptation to changing tissue or geographic environments. In addition to mtDNA sequence variation, mtDNA copy number and perhaps transfer of mtDNA sequences into the nucleus can contribute to certain cancers. Strong functional relevance of mtDNA variation has been demonstrated in oncocytoma and prostate cancer, while mtDNA variation has been reported in multiple other cancer types. Alterations in nuclear DNA-encoded mitochondrial genes have confirmed the importance of mitochondrial metabolism in cancer, affecting mitochondrial reactive oxygen species production, redox state and mitochondrial intermediates that act as substrates for chromatin-modifying enzymes. Hence, subtle changes in the mitochondrial genotype can have profound effects on the nucleus, as well as carcinogenesis and cancer progression.

Lactate Limits T Cell Proliferation via the NAD(H) Redox State.

Immune cell function is influenced by metabolic conditions. Low-glucose, high-lactate environments, such as the placenta, gastrointestinal tract, and the tumor microenvironment, are immunosuppressive, especially for glycolysis-dependent effector T cells. We report that nicotinamide adenine dinucleotide (NAD+), which is reduced to NADH by lactate dehydrogenase in lactate-rich conditions, is a key point of metabolic control in T cells. Reduced NADH is not available for NAD+-dependent enzymatic reactions involving glyceraldehyde 3-phosphate dehydrogenase (GAPDH) and 3-phosphoglycerate dehydrogenase (PGDH). We show that increased lactate leads to a block at GAPDH and PGDH, leading to the depletion of post-GAPDH glycolytic intermediates, as well as the 3-phosphoglycerate derivative serine that is known to be important for T cell proliferation. Supplementing serine rescues the ability of T cells to proliferate in the presence of lactate-induced reductive stress. Directly targeting the redox state may be a useful approach for developing novel immunotherapies in cancer and therapeutic immunosuppression.

Foxp3 Reprograms T Cell Metabolism to Function in Low-Glucose, High-Lactate Environments.

Immune cells function in diverse metabolic environments. Tissues with low glucose and high lactate concentrations, such as the intestinal tract or ischemic tissues, frequently require immune responses to be more pro-tolerant, avoiding unwanted reactions against self-antigens or commensal bacteria. T-regulatory cells (Tregs) maintain peripheral tolerance, but how Tregs function in low-glucose, lactate-rich environments is unknown. We report that the Treg transcription factor Foxp3 reprograms T cell metabolism by suppressing Myc and glycolysis, enhancing oxidative phosphorylation, and increasing nicotinamide adenine dinucleotide oxidation. These adaptations allow Tregs a metabolic advantage in low-glucose, lactate-rich environments; they resist lactate-mediated suppression of T cell function and proliferation. This metabolic phenotype may explain how Tregs promote peripheral immune tolerance during tissue injury but also how cancer cells evade immune destruction in the tumor microenvironment. Understanding Treg metabolism may therefore lead to novel approaches for selective immune modulation in cancer and autoimmune diseases.

Targeting ACLY sensitizes castration-resistant prostate cancer cells to AR antagonism by impinging on an ACLY-AMPK-AR feedback mechanism.

The androgen receptor (AR) plays a central role in prostate tumor growth. Inappropriate reactivation of the AR after androgen deprivation therapy promotes development of incurable castration-resistant prostate cancer (CRPC). In this study, we provide evidence that metabolic features of prostate cancer cells can be exploited to sensitize CRPC cells to AR antagonism. We identify a feedback loop between ATP-citrate lyase (ACLY)-dependent fatty acid synthesis, AMPK, and the AR in prostate cancer cells that could contribute to therapeutic resistance by maintaining AR levels. When combined with an AR antagonist, ACLY inhibition in CRPC cells promotes energetic stress and AMPK activation, resulting in further suppression of AR levels and target gene expression, inhibition of proliferation, and apoptosis. Supplying exogenous fatty acids can restore energetic homeostasis; however, this rescue does not occur through increased ß-oxidation to support mitochondrial ATP production. Instead, concurrent inhibition of ACLY and AR may drive excess ATP consumption as cells attempt to cope with endoplasmic reticulum (ER) stress, which is prevented by fatty acid supplementation. Thus, fatty acid metabolism plays a key role in coordinating ER and energetic homeostasis in CRPC cells, thereby sustaining AR action and promoting proliferation. Consistent with a role for fatty acid metabolism in sustaining AR levels in prostate cancer in vivo, AR mRNA levels in human prostate tumors correlate positively with expression of ACLY and other fatty acid synthesis genes. The ACLY-AMPK-AR network can be exploited to sensitize CRPC cells to AR antagonism, suggesting novel therapeutic opportunities for prostate cancer.

Mitochondrial functions modulate neuroendocrine, metabolic, inflammatory, and transcriptional responses to acute psychological stress.

The experience of psychological stress triggers neuroendocrine, inflammatory, metabolic, and transcriptional perturbations that ultimately predispose to disease. However, the subcellular determinants of this integrated, multisystemic stress response have not been defined. Central to stress adaptation is cellular energetics, involving mitochondrial energy production and oxidative stress. We therefore hypothesized that abnormal mitochondrial functions would differentially modulate the organism's multisystemic response to psychological stress. By mutating or deleting mitochondrial genes encoded in the mtDNA [NADH dehydrogenase 6 (ND6) and cytochrome c oxidase subunit I (COI)] or nuclear DNA [adenine nucleotide translocator 1 (ANT1) and nicotinamide nucleotide transhydrogenase (NNT)], we selectively impaired mitochondrial respiratory chain function, energy exchange, and mitochondrial redox balance in mice. The resulting impact on physiological reactivity and recovery from restraint stress were then characterized. We show that mitochondrial dysfunctions altered the hypothalamic-pituitary-adrenal axis, sympathetic adrenal-medullary activation and catecholamine levels, the inflammatory cytokine IL-6, circulating metabolites, and hippocampal gene expression responses to stress. Each mitochondrial defect generated a distinct whole-body stress-response signature. These results demonstrate the role of mitochondrial energetics and redox balance as modulators of key pathophysiological perturbations previously linked to disease. This work establishes mitochondria as stress-response modulators, with implications for understanding the mechanisms of stress pathophysiology and mitochondrial diseases.

Rac2-MRC-cIII-generated ROS cause genomic instability in chronic myeloid leukemia stem cells and primitive progenitors.

Chronic myeloid leukemia in chronic phase (CML-CP) is induced by BCR-ABL1 oncogenic tyrosine kinase. Tyrosine kinase inhibitors eliminate the bulk of CML-CP cells, but fail to eradicate leukemia stem cells (LSCs) and leukemia progenitor cells (LPCs) displaying innate and acquired resistance, respectively. These cells may accumulate genomic instability, leading to disease relapse and/or malignant progression to a fatal blast phase. In the present study, we show that Rac2 GTPase alters mitochondrial membrane potential and electron flow through the mitochondrial respiratory chain complex III (MRC-cIII), thereby generating high levels of reactive oxygen species (ROS) in CML-CP LSCs and primitive LPCs. MRC-cIII-generated ROS promote oxidative DNA damage to trigger genomic instability, resulting in an accumulation of chromosomal aberrations and tyrosine kinase inhibitor-resistant BCR-ABL1 mutants. JAK2(V617F) and FLT3(ITD)-positive polycythemia vera cells and acute myeloid leukemia cells also produce ROS via MRC-cIII. In the present study, inhibition of Rac2 by genetic deletion or a small-molecule inhibitor and down-regulation of mitochondrial ROS by disruption of MRC-cIII, expression of mitochondria-targeted catalase, or addition of ROS-scavenging mitochondria-targeted peptide aptamer reduced genomic instability. We postulate that the Rac2-MRC-cIII pathway triggers ROS-mediated genomic instability in LSCs and primitive LPCs, which could be targeted to prevent the relapse and malignant progression of CML.