Age-related macular degeneration (AMD), the leading cause of vision loss in the elderly, is a degenerative disease of the macula, where retinal pigment epithelium (RPE) cells are damaged in the early stages of the disease, and chronic inflammatory processes may be involved. Besides aging and lifestyle factors as drivers of AMD, a strong genetic association to AMD is found in genes of the complement system, with a single polymorphism in the complement factor H gene (CFH), accounting for the majority of AMD risk. However, the exact mechanism of CFH dysregulation confers such a great risk for AMD and its role in RPE cell homeostasis is unclear. To explore the role of endogenous CFH locally in RPE cells, we silenced CFH in human hTERT-RPE1 cells. We demonstrate that endogenously expressed CFH in RPE cells modulates inflammatory cytokine production and complement regulation, independent of external complement sources, or stressors. We show that loss of the factor H protein (FH) results in increased levels of inflammatory mediators (e.g., IL-6, IL-8, GM-CSF) and altered levels of complement proteins (e.g., C3, CFB upregulation, and C5 downregulation) that are known to play a role in AMD. Moreover, our results identify the NF-κB pathway as the major pathway involved in regulating these inflammatory and complement factors. Our findings suggest that in RPE cells, FH and the NF-κB pathway work in synergy to maintain inflammatory and complement balance, and in case either one of them is dysregulated, the RPE microenvironment changes towards a proinflammatory AMD-like phenotype.

Age-related macular degeneration (AMD) is a complex degenerative disease of the retina with multiple risk-modifying factors, including aging, genetics, and lifestyle choices. The combination of these factors leads to oxidative stress, inflammation, and metabolic failure in the retinal pigment epithelium (RPE) with subsequent degeneration of photoreceptors in the retina. The alternative complement pathway is tightly linked to AMD. In particular, the genetic variant in the complement factor H gene (CFH), which leads to the Y402H polymorphism in the factor H protein (FH), confers the second highest risk for the development and progression of AMD. Although the association between the FH Y402H variant and increased complement system activation is known, recent studies have uncovered novel FH functions not tied to this activity and highlighted functional relevance for intracellular FH. In our previous studies, we show that loss of CFH expression in RPE cells causes profound disturbances in cellular metabolism, increases the vulnerability towards oxidative stress, and modulates the activation of pro-inflammatory signaling pathways, most importantly the NF-kB pathway. Here, we silenced CFH in hTERT-RPE1 cells to investigate the mechanism by which intracellular FH regulates RPE cell homeostasis. We found that silencing of CFH results in hyperactivation of mTOR signaling along with decreased mitochondrial respiration and that mTOR inhibition via rapamycin can partially rescue these metabolic defects. To obtain mechanistic insight into the function of intracellular FH in hTERT-RPE1 cells, we analyzed the interactome of FH via immunoprecipitation followed by mass spectrometry-based analysis. We found that FH interacts with essential components of the ubiquitin-proteasomal pathway (UPS) as well as with factors associated with RB1/E2F signalling in a complement-pathway independent manner. Moreover, we found that FH silencing affects mRNA levels of the E3 Ubiquitin-Protein Ligase Parkin and PTEN induced putative kinase (Pink1), both of which are associated with UPS. As inhibition of mTORC1 was previously shown to result in increased overall protein degradation via UPS and as FH mRNA and protein levels were shown to be affected by inhibition of UPS, our data stress a potential regulatory link between endogenous FH activity and the UPS.

Despite comprehensive research efforts over the last decades, the pathomechanisms of age-related macular degeneration (AMD) remain far from being understood. Large-scale genome wide association studies (GWAS) were able to provide a defined set of genetic aberrations which contribute to disease risk, with the strongest contributors mapping to distinct regions on chromosome 1 and 10. While the chromosome 1 locus comprises factors of the complement system with well-known functions, the role of the 10q26-locus in AMD-pathophysiology remains enigmatic. 10q26 harbors a cluster of three functional genes, namely PLEKHA1, ARMS2 and HTRA1, with most of the AMD-associated genetic variants mapping to the latter two genes. High linkage disequilibrium between ARMS2 and HTRA1 has kept association studies from reliably defining the risk-causing gene for long and only very recently the genetic risk region has been narrowed to ARMS2, suggesting that this is the true AMD gene at this locus. However, genetic associations alone do not suffice to prove causality and one or more of the 14 SNPs on this haplotype may be involved in long-range control of gene expression, leaving HTRA1 and PLEKHA1 still suspects in the pathogenic pathway. Both, ARMS2 and HTRA1 have been linked to extracellular matrix homeostasis, yet their exact molecular function as well as their role in AMD pathogenesis remains to be uncovered. The transcriptional regulation of the 10q26 locus adds an additional level of complexity, given, that gene-regulatory as well as epigenetic alterations may influence expression levels from 10q26 in diseased individuals. Here, we provide a comprehensive overview on the 10q26 locus and its three gene products on various levels of biological complexity and discuss current and future research strategies to shed light on one of the remaining enigmatic spots in the AMD landscape.

Advances in artificial intelligence (AI) promise to reshape the landscape of scientific inquiry. Amidst all these, OpenAI’s latest tool, Deep Research, stands out for its potential to revolutionize how researchers engage with the literature. However, this leap forward presents a paradox; while AI-generated reviews offer speed and accessibility with minimal effort, they raise fundamental concerns about citation integrity, critical appraisal, and the erosion of deep scientific thinking. These concerns are particularly problematic in the context of biomedical research, where evidence quality may influence clinical practice and decision-making. In this piece, we present an empirical evaluation of Deep Research and explore both its remarkable capabilities and inherent limitations. Through structured experimentation, we assess its effectiveness in synthesizing literature, highlight key shortcomings, and reflect on the broader implications of these tools for research training, and the integrity of evidence-based practice. With AI tools increasingly blurring the lines between knowledge generation and critical inquiry, we argue that while AI democratizes access to knowledge, wisdom remains distinctly human.

This study aims to evaluate whether in-context learning (ICL), a prompt-based learning mechanism enabling multimodal foundation models to rapidly adapt to new tasks without retraining or large annotated datasets, can achieve comparable diagnostic performance to domain-specific foundation models. Specifically, we use diabetic retinopathy (DR) detection as an exemplar task to probe if a multimodal foundation model (Google Gemini 1.5 Pro), employing ICL, can match the performance of a domain-specific model (RETFound) fine-tuned explicitly for DR detection from color fundus photographs (CFPs). A cross-sectional study. A retrospective, publicly available dataset (Indian Diabetic Retinopathy Image Dataset) comprising 516 CFPs collected at an eye clinic in India, featuring both healthy individuals and patients with DR. The images were dichotomized into 2 groups based on the presence or absence of any signs of DR. RETFound was fine-tuned for this binary classification task, while Gemini 1.5 Pro was assessed for it under zero-shot and few-shot prompting scenarios, with the latter incorporating random or k-nearest-neighbors-based sampling of a varying number of example images. For experiments, data were partitioned into training, validation, and test sets in a stratified manner, with the process repeated for 10-fold cross-validation. Performance was assessed via accuracy, F1 score, and expected calibration error of predictive probabilities. Statistical significance was evaluated using Wilcoxon tests. The best ICL performance with Gemini 1.5 Pro yielded an average accuracy of 0.841 (95% confidence interval [CI]: 0.803-0.879), an F1 score of 0.876 (95% CI: 0.844-0.909), and a calibration error of 0.129 (95% CI: 0.107-0.152). RETFound achieved an average accuracy of 0.849 (95% CI: 0.813-0.885), an F1 score of 0.883 (95% CI: 0.852-0.915), and a calibration error of 0.081 (95% CI: 0.066-0.097). While accuracy and F1 scores were comparable (P > 0.3), RETFound's calibration was superior (P = 0.004). Gemini 1.5 Pro with ICL demonstrated performance comparable to RETFound for binary DR detection, illustrating how future medical artificial intelligence systems may build upon such frontier models rather than being bespoke solutions. Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.