Finding. The human AMD risk protein ARMS2 A69S, dropped into the mouse genome next to Htra1, is enough to wound retinal mitochondria early and the outer retina later. Gogna, Nishina and colleagues insert the complete human ARMS2 gene, risk or wild-type, 3.5 kb upstream of mouse Htra1, promoter and orientation intact. Rodents have no ARMS2 of their own. Expression lands where humans put it: neural tissue, testis, eye, and inside the eye in horizontal cells and retinal pigment epithelium. A69S does not change how much RNA is made. By 12 months the risk mice have more fundus spots, a thinner outer nuclear layer the wild-type human gene does not cause, and smaller scotopic ERGs. At postnatal day 14, before that, OXPHOS genes are already down in retina and RPE, retinal mtDNA copy number is down, and maximal respiration is down. Sphingolipids in the RPE may be remodeling.
Why this paper matters
10q26 is the strongest common AMD locus, and ARMS2 versus HTRA1 has been a linkage-disequilibrium stalemate because mice lack ARMS2. A humanized pair that keeps Htra1 spacing and still gives an A69S-specific ONL phenotype is a genetic argument. A P14 mitochondrial package (transcripts, copy number, max respiration) is the organelle argument. Together they say: this risk allele is not waiting until seventy to touch energy metabolism.
What they actually measured
Two knock-ins, expression atlas, 12-month eye phenotyping, P14 multi-omic/mitochondrial assays. Htra1 neighborhood is designed, not ignored.
How to read the score
High 80s. Humanized AMD allele plus early mitochondrial function. Confidence is high for the mouse phenotypes as stated.
Caveats
Not the full human haplotype. Preliminary lipids. Early mito versus late structure is a timeline, not a proven causal chain.
What to do with it
If you test AMD metabolic therapies, this is the ARMS2 animal. Pull P14 respiration and 12-month ERG as the two ends of the assay. Do not drop HTRA1 from the human genetic story; this model isolates ARMS2, it does not erase the neighbor.