🧬 PubMed RNA Editing Daily Digest

Understanding transcriptional regulation of native genomic elements requires tools that combine high sensitivity, quantitative output, and broad applicability. Existing methods often have limited dynamic range, disrupt host RNAs, or fail to detect short-lived transcripts. Here, we present ribozyme-processed ADAR-engaging RNA-directed editing (REDDIT), a technology that converts transcriptional events into reporter protein translation via precise A-to-I RNA editing. REDDIT sensitively detects transcription from protein-coding genes, long noncoding RNAs (lncRNAs), primary microRNAs (pri-miRNAs), and enhancer RNAs (eRNAs), including low abundance and short-lived species, while minimally perturbing host gene expression, RNA processing, and the global editome. We apply REDDIT to monitor the naïve-to-primed transition in human embryonic stem cells (hESCs) and convert it into a transcription recorder that permanently logs transient and combinatorial transcriptional inputs when paired with Cre recombinase. Finally, by adapting REDDIT for high-throughput screening, we uncover multiple signaling pathways that regulate lncRNA and eRNA biogenesis. REDDIT therefore provides a scalable platform for quantitative monitoring and retrospective analysis of endogenous transcriptional dynamics across diverse genomic contexts.

Drug resistance in patients remains a significant obstacle to successful treatment, even with improvements in cancer treatment strategies. Resistance to taxanes, such as docetaxel (Dtx) and cabazitaxel (Cbz), frequently emerges in castration resistant prostate cancer (CRPC). Through pulse selection of the parental cells (DU145), we established Dtx- and Cbz-resistant CRPC cell models and integrated different omic approaches, including transcriptomics and proteomics, to determine the molecular signatures underlying taxane resistance. Interestingly, several genes were regulated in the same direction (up- or down-regulation) at both the gene and protein expression levels in resistant cells compared to parental cells, suggesting that alterations primarily occur at the transcriptional level and manifest at the protein level. Among the differentially regulated genes, Cysteine Rich Protein 2 (CRIP2), a gene associated with tumor suppressor function, has been found to be the most downregulated in taxane-resistant cells. Conversely, Nicotinamide N-Methyltransferase (NNMT) exhibited a significant upregulation and has been validated in the context of taxane resistance. Its overexpression was shown to promote taxane resistance in two different CRPC cell lines, whereas depletion via siRNA or gRNA, as well as treatment with 1-methylnicotinamide (1-MNA, used as a feedback inhibitor)resensitized the resistant cells. RNA-sequencing of NNMT-knockout (CRISPR-Cas9) cells has indicated involvement of TGFβ signaling, and suppressing this pathway has further increased the taxane sensitivity. Epithelial Mesenchymal Transition (EMT) was another pathway depleted upon knockout, and subsequent analysis revealed a significant correlation between NNMT and EMT-related genes (VIM, CDH2, FN1, TGFB1, and ZEB2) in both the Cancer Cell Line Encyclopedia (CCLE) panel and patient data. Additionally, in cancers other than PC, NNMT has been observed to predict treatment outcomes, and notably, among the patients with a high EMT signature, elevated NNMT levels were associated with decreased overall survival. More importantly, NNMT-high patients were found to be non-responders to taxane-containing chemotherapy regimens. Collectively, our findings suggest that targeting NNMT and the pathways it affects, such as TGFβ, offers a viable approach for addressing taxane-resistant PC.

The convergence of genome editing (GE) and artificial intelligence (AI) is shifting crop improvement from empirical optimization toward predictive design. In this Perspective, we propose that GE and AI are linked by a reciprocal innovation cycle. AI improves GE by enabling more accurate guide RNA design, editing outcome and off-target prediction, and data-driven development of next-generation editing systems. In turn, GE provides a powerful experimental platform to validate AI predictions, dissect causal genotype-to-phenotype relationships, and generate high-resolution perturbation datasets for iterative model refinement. This bidirectional interplay is beginning to accelerate the engineering of complex agronomic traits, including trait stacking, metabolic rewiring, climate resilience, stress tolerance, nutritional enhancement, and de novo domestication. Here, we argue that this synergy is emerging most clearly in predictive editing design and iterative model validation, although robust breeding-scale evidence remains limited. We also discuss emerging opportunities for AI-guided closed-loop editing platforms and generative biological design, together with key bottlenecks, including limited training data, biological complexity across scales, model interpretability, and heterogeneous regulatory landscapes. We argue that integrating AI with GE can establish a practical framework for predictive crop design and enable more efficient and sustainable crop engineering.