Genome-wide studies have found dozens of risk loci for age-related macular degeneration, but for many of them the gene doing the work is unknown. This study combines AMD genetics with blood and retinal gene-expression data to prioritize nine candidate genes, screens them in zebrafish, and follows the strongest, CNN2, into knockout mice. Mice lacking Cnn2 lose sensitivity in dim light and to fine contrast while acuity stays normal, and their photoreceptor layer thins, even though the gene is expressed mainly in the retina's blood-vessel cells rather than in photoreceptors.
0:00Welcome to Base by Base, the papercast that brings genomics to you, wherever you are. Thanks for listening, and don't forget to follow and rate us in your podcast app. Base by Base is now on YouTube too, at Base by Base, where every episode gets a video with chapters and the full description. Come subscribe.
0:21Picture a kitchen at dusk. The sun has gone, and nobody has switched on a lamp yet. Most of us barely notice that half hour. We still find the kettle, and we still read the clock on the wall. So what is the eye doing in that moment? It is pulling a picture out of very little light, and telling apart greys that are almost the same. Now imagine that skill slowly wearing down. In early age-related macular degeneration, sensitivity to fine contrast tends to fade before sharpness does. You can still read the letters on the chart. You just start losing the subtle shades.
1:03Here is one way to think about the retina. Picture a stage in a theater. The performers are the cells that catch light, the rods and the cones, and they are the ones the audience watches. But no show survives without the crew backstage, the people who keep the power running, the air moving and the lights in place. You never see the crew during the play. You only notice them when something quietly goes wrong. Most of what we know about eye disease is written about the performers. Today's story is about a gene that seems to work with the crew.
1:36Macular degeneration is one of the best mapped complex eye diseases in human genetics. Huge studies have pinned down dozens of regions of the genome that raise the risk. But a region of the genome is not an answer in itself. Each one can hold several genes, and for many of them nobody knows which gene is actually responsible. So how do you pick the right gene out of a crowded stretch of DNA? And once you think you have found it, what happens to vision when you take that gene away?
2:08Today, we celebrate the work of Fei-Fei Cheng, Hao Mou and colleagues, led by Jian Yang and Zi-Bing Jin, from Westlake University in Hangzhou and the Beijing Institute of Ophthalmology at Capital Medical University, with partners in Wenzhou, Brisbane and Sydney, who have advanced our understanding of how genetic risk for macular degeneration reaches the retina. Their paper, Genetic evidence and cross-species functional characterization implicate C N N two in age-related macular degeneration susceptibility, was published in the Proceedings of the National Academy of Sciences in September 2026.
2:47Let's start with the disease itself. Age-related macular degeneration damages the macula, the small central patch of the retina we use for detail. That is the part you read with, and the part you use to recognize a face across a room. As it fails, central vision declines, and the loss cannot be reversed. How common is it? Among people aged forty-five to eighty-five, the global prevalence sits at eight point six nine percent. By 2040, the number of people affected worldwide is projected to reach 288 million.
3:23What drives the disease? Smoking, nutrition and cardiovascular disease all have a significant impact on how it progresses. And genetics matters a great deal. Macular degeneration is one of the most genetically well-defined complex eye diseases we know. Genome-wide association studies, which compare the genomes of many thousands of people, have found at least sixty-three regions linked to its risk. Together, those regions explain over half of the heritability of susceptibility. For a common disease of aging, that is an unusually strong genetic map to work from.
3:59So where is the gap? A risk region is a stretch of DNA, not an explanation. It often holds several genes sitting side by side. And the variant that raises risk may not sit inside a gene at all. It may sit in a switch that turns a nearby gene up or down. For many of these regions, nobody knows which gene is the real target. Worse, most of the genes that have been nominated as candidates have never been tested in a living animal. That is the bottleneck this study sets out to tackle.
4:32Why does that bottleneck matter so much? Because you cannot understand a disease, or design a treatment, around a bare stretch of DNA. You need a gene, a cell type and a mechanism. Earlier work had already shown what animal models can add. A zebrafish model helped explain a rare risk variant in the gene C F I, and two-year-old mice lacking the gene C F H showed visual problems. But testing every candidate that way is slow. The field needed an efficient and affordable way to screen.
5:08Here is how the team narrowed the field. They used a method called summary-data-based Mendelian randomization. The idea is simpler than the name. If a genetic variant raises disease risk, and the same variant also changes how much of a nearby gene gets made, then that gene becomes a strong candidate. Gene activity becomes the bridge between the variant and the disease. To run it, they combined two large datasets. The first was a genome-wide study of sixteen thousand one hundred and forty-four people with the disease and seventeen thousand eight hundred and thirty-two controls.
5:46The second dataset measured gene activity in the blood of two thousand seven hundred and sixty-five people. Crossing the two, the team found sixteen genes whose activity tracked with disease risk. Then came a check called HEIDI, and it asks a sharp question. Is one variant really doing both jobs, or are two different variants simply sitting close together? Only genes that passed that test survived, and that left nine. There is an obvious objection here, though. Blood is not the eye, so why should blood tell us anything about the retina?
6:22So the team went looking in the eye itself. They repeated the analysis with gene activity measured in retinas from four hundred and six donors. Four of the nine genes held up, with the same direction of effect. For a dataset that small, that is a solid rate of replication. They also checked forty-eight other human tissues, and every one of the nine genes showed a signal in at least two of them. Put another way, these genes are not eye-only curiosities. Part of the risk may run through the whole body.
6:56Statistics can point, but only biology can confirm. So the next step was a living animal, and the team chose zebrafish larvae, which are widely used to model eye disorders and whose eyes are easy to measure. Four of the nine genes had zebrafish counterparts with more than sixty percent similarity. The team switched each one down with morpholinos, short synthetic molecules that stop a gene's message from being turned into protein. Then came the crucial control. They added the gene's message back, to see whether the damage reversed.
7:32Genes that passed the fish screen moved to mice, where a mammalian retina can be examined in far more detail. For C N N two, the team built knockout mice with CRISPR, removing the gene entirely. How do you ask a mouse what it can see? One way is an electroretinogram, which records the retina's electrical response to flashes, both in bright light and in very dim light. Another is an optokinetic test with rotating stripes. A mouse that sees the stripes turns its head to follow them, so you make them finer or fainter until it stops.
8:09So what happened in the fish? Two of the four genes mattered. Knocking down C N N two shrank the eye. Eye area fell by forty-eight point five percent, and the length of the eyeball fell by thirty-two point five percent. The second gene, S A R M one, did the same on a smaller scale, cutting eye area by twenty-eight point five percent. The other two genes made no significant difference at all. And the rescue worked. When the team put the C N N two message back, eye length recovered by ninety-one percent.
8:45Smaller eyes are one thing, but could the fish still see? The team tested that with a visual motor response. Healthy larvae jolt into motion when the lights switch on or off. Larvae with C N N two knocked down reacted weakly and late. Their peak response to the lights coming on dropped by sixty-one point nine percent. Twenty seconds after the lights went off, their activity was seventy-eight point six percent lower than in controls. Adding the gene's message back brought part of the response back, though not all of it.
9:18Here the story narrows to a single gene. Mice lacking S A R M one showed little or no visible problem at four and a half months, so the team set them aside. C N N two, meanwhile, kept collecting evidence. Several complementary statistical methods pointed to it. And a fine-mapped risk variant sat right in its promoter, the stretch of DNA that switches the gene on, inside a region that is open and active in the tissue behind the retina. Why does that matter? Because it is exactly where a variant could turn the gene up or down.
9:56So what does a mouse without C N N two actually see? In bright light, its retina responded normally. In dim light, it did not. At a dim flash of light, the first wave of the electrical response fell by thirty-nine point two percent, and the second wave fell by twenty-seven point three percent. At brighter flashes, the difference disappeared. The authors read this as a modest loss of sensitivity in low light. In preliminary tests, the dip was already visible as early as one month of age.
10:29The stripe test told a similar story. Visual acuity, the ability to resolve fine stripes, was no different from that of normal mice. Contrast was another matter. At one stripe setting, normal mice noticed the pattern at four point one percent contrast. Mice without C N N two needed fifteen point three percent. At a finer setting, the thresholds were twenty-nine point one percent against fifty-one point eight percent. The mice could still see the pattern, they just needed it much bolder. And that mirrors early macular degeneration in people, where fine contrast fades before sharpness.
11:07The structure of the retina changed too. Scans showed the photoreceptor layer thinning at three months and again at five months of age. Staining revealed fewer cone cells marked by cone arrestin, and a sparser pattern of rhodopsin, a rod marker. Protein measurements confirmed significant drops in both. A third marker, recoverin, trended downward without reaching significance. Across the whole retina, four hundred and thirty-six genes changed their activity, and those genes were enriched six point zero two fold in regions already linked to the disease by human genetic studies.
11:45Now comes the twist. Where is C N N two actually active? In the wild-type mouse retina, the protein sat alongside a marker of the cells that line blood vessels. Single-cell data told the same story, with the gene expressed mostly in endothelial cells and in pericytes, two cell types of the retina's vessels. In the rods and cones themselves, its expression was minimal. Back to our stage for a moment. The performers are the ones fading, yet the gene belongs mostly to the crew. And in the knockout retina, the inferred numbers of endothelial cells, pericytes and rods were all lower.
12:26What is this gene, exactly? C N N two makes calponin two, a protein that binds the actin skeleton inside cells. In other settings it has been tied to cell movement, to the migration of cells that form new vessels, and to inflammation. The authors propose a chain of events from there. Losing calponin two may disturb the balance between the retina's vessels and its neurons, and the light-catching cells may suffer as a consequence. It is a plausible and interesting idea. But it is a proposal, not a demonstration.
13:01And the authors say so plainly. Their data do not establish whether the damage to the photoreceptors comes directly from losing the gene, or indirectly through the vessels or other cells. Sorting that out will take experiments that remove the gene from one cell type at a time. There is a second caution, and it matters a lot. In donated human retinas, C N N two activity was generally lower in people with the disease, especially in the macula. But the difference was not statistically significant.
13:34Then there are the animal models. Neither zebrafish nor mice have a macula, the very spot the disease attacks in people, and no prominent drusen appeared within the time frame of the study. The fish were larvae, so the effects of aging could not be studied at all. The mice were followed to five months, and the authors call for studies out to twelve to eighteen months. And removing a gene completely is a far stronger hit than a common risk variant, which usually nudges a gene's activity only slightly.
14:06One more limit sits in the genetics itself. The discovery step relied on blood, because blood datasets are large, and that choice buys statistical power. But it can miss genes whose effect on risk shows up only in the retina. So why does this study still matter? Because it turns a statistical signal into biology you can test, and it offers a pipeline that other complex diseases can reuse. It also leaves an intriguing loose end. C N N two sits in a region linked to late-onset Alzheimer's disease as well, hinting at possible shared biology.
14:45So here is where this leaves us. Starting from dozens of genetic risk regions, this team narrowed the list to nine genes, tested four of them in fish, and followed one into mice. Removing C N N two dimmed the retina's response in low light, blunted sensitivity to fine contrast while leaving sharpness intact, and thinned the photoreceptor layer. And yet the gene lives mostly in the cells of the retina's vessels. In our stage picture, a member of the crew went missing, and the performers faltered in the dimmest scenes.
15:20Whether that missing crew member starves the performers, or harms them in some other way, is still an open question. So is whether the same thing happens in human eyes. But the next time the light fades in a quiet kitchen, it may be worth remembering that seeing at dusk could depend on more than the cells that catch the light. What does this mean for the future of treatments that look beyond the photoreceptors, to the vessels that keep them running?
15:48This episode was based on an open access article under the C C B Y four point zero license. You can find a direct link to the paper and the license in our episode description. If you enjoyed this, follow or subscribe in your podcast app and leave a five-star rating. If you'd like to support our work, use the donation link in the description. Now, stay with us for an original track created especially for this episode and inspired by the article you've just heard about. Thanks for listening, and join us next time as we explore more science, base by base.
