Feedback regulation between ALKBH3-mediated glycolysis and histone lactylation promotes age-related macular degeneration

Age-related macular degeneration (AMD) is a leading cause of blindness among the elderly. Using clinical samples and knockout mice, we reported that the m1A eraser ALKBH3 reshaped retinal metabolism to promote AMD. In retinal pigment epithelium (RPE), the dm1ACRISPR system demonstrated that ALKBH3 demethylated the glycolytic enzyme HK2 to activate anaerobic glycolysis, producing excessive lactate. The lactate promoted histone lactylation at H3K18, which in turn bound to ALKBH3 to amplify its transcription, establishing a positive feedback loop. The ALKBH3 inhibitor HUHS015 disrupted this loop, effectively mitigating RPE degeneration. Furthermore, ALKBH3 directly targeted the pro-angiogenic factor VEGFA to modulate the metabolic cross-talk between RPE and choroidal capillaries, thus promoting choroidal neovascularization (CNV). HUHS015 inhibited CNV synergistically with the anti-VEGF drug Aflibercept. Our study provides critical insights into the molecular mechanisms and metabolic events facilitating the progression from RPE degeneration to CNV in AMD, laying the groundwork for new treatments of AMD. Extracted and purified total RNA was evaluated for quantity and integrity. Ideal RNA samples were sent for library construction using the Directional RNA Library Prep Kit (GenSeq, Shanghai, China), according to the manufacturer’s instructions. The mRNA is typically fragmented into pieces of ~300 nucleotides in length. cDNA synthesis form the mRNA fragments was then carried out using reverse transcriptase and random hexamer primers. The resulting cDNA fragments were subjected to end-repair and dA-tailing, followed by adapter ligation. These adapter-ligated cDNA samples were subsequently amplified via PCR and purified to yield sequencing libraries. Finally, the libraries were sequenced on an Illumina Novaseq 6000 platform (Illumina) using a paired-end 150 bp mode. Raw sequencing data were processed and filtered to produce clean reads, which were aligned to the human genome using the STAR software. To generate expression levels, we used the HTSeq v0.6.0 to calculate the fragments per kilobase per million mapped reads (FPKM) of each gene or transcript. The DESeq2 algorithm was applied to identify differentially expressed genes.