SMARCAD1 ATPase activity is required to silence endogenous retroviruses in embryonic stem cells

Picture the earliest moments of mammalian life. The genome is wide open, buzzing, and permissive. That’s great for building a body from scratch, but it comes with a problem: stowaways in the DNA—endogenous retroviruses, or ERVs—are waiting for a chance to speak. If they turn on at the wrong time or place, they can hijack nearby genes and scramble development. In humans and mice, these retroviral fossils are not fringe DNA; they’re a solid slice of the genome. Roughly a tenth of our sequence traces back to them. So, embryonic stem cells live on a knife edge: they need a general openness to make any cell type, yet they must carve out pockets of deep silence to keep these elements asleep. We’ve known for years how cells start that silence. A family of zinc-finger proteins recognizes specific retroviral sequences and brings in KAP1—also called TRIM28—a scaffold that recruits the histone methyltransferase SETDB1. SETDB1 lays down the H3K9me3 mark, and together they build heterochromatin—tightly packed, repressive chromatin—over retroviral promoters and their regulatory sequences. That’s the canonical story: KRAB zinc-finger proteins, KAP1, SETDB1, and H3K9me3. Clean and linear. But there’s been a missing link. Silence isn’t just about marks; it often needs energy—literal ATP—to move nucleosomes and remodel chromatin. Where does ATP-dependent remodeling fit into this safeguard? And if it matters, which remodeler does the work? Sachs, Ding, Bergmaier, and colleagues went after that question and homed in on SMARCAD1. It’s a SWI-SNF-like chromatin remodeler with an ATPase motor and a track record of working on compacted DNA. Proteomics had hinted at a partnership between SMARCAD1 and KAP1 in mouse embryonic stem cells, and SMARCAD1 had been spotted at retroelements that carry repressive marks. The hunch was simple but bold: maybe KAP1 brings SMARCAD1 to retroviral DNA, and SMARCAD1’s motor helps lock the whole repressive package in place. They started by asking where SMARCAD1 sits in the genome of embryonic stem cells and whether it actually favors retroviral territory. To get clean maps, they tuned the chemistry—dual crosslinking with DSG and formaldehyde—and paired that with both a FLAG-tagged SMARCAD1 line and the endogenous protein. With that setup, they called five thousand seven hundred twenty-seven regions enriched for SMARCAD1, and when they overlapped the FLAG and endogenous datasets, they landed on two thousand three hundred eighty high-confidence sites. Most of those sites were not in gene bodies. About seventy-two percent fell in intergenic space, exactly where you’d expect transposons to lurk. What really mattered, though, was who SMARCAD1 was keeping company with. Across the genome, SMARCAD1 showed up in places already stamped as repressive: H3K9me3 and another heterochromatin mark, H4K20me3. A striking fraction of SMARCAD1 peaks overlapped H3K9me3—about eighty-five percent—and an even higher share lined up with KAP1—around eighty-seven percent. That’s not a casual encounter. In fractionation experiments, SMARCAD1 traveled with KAP1 and SETDB1 in big protein assemblies, and in nuclear extracts, SMARCAD1 pulled down SETDB1 and HDAC1, the deacetylase that helps tighten chromatin. This looked like a single team, not a coincidental crowd. Zooming in on transposons, the pattern sharpened. SMARCAD1 parked on long terminal repeat retrotransposons, especially class I and class II families. The stars of the show were IAPs—intracisternal A-particles—some of the most active elements in mouse. SMARCAD1’s footprint was strongest over the five-prime untranslated region of IAPs, a hotspot for promoter and enhancer activity in these elements. It also appeared on VL30 and murine leukemia virus family members, both retrovirus-like. But two things stood out: it did not pile onto MERVL, a class III element with a different regulatory life, and it largely ignored LINEs—long interspersed nuclear elements—even though LINEs make up about a fifth of the mouse genome. In re-ChIP assays—essentially doing chromatin immunoprecipitation twice on the same DNA fragments—SMARCAD1 and KAP1 were found together on the same retroviral copies, including specific loci like Mier3. So, the co-occupancy wasn’t just averaged across families; it was simultaneous at single elements. Position is one thing. Function is another. When they depleted SMARCAD1, the repressive landscape over IAPs changed. KAP1 binding dipped. H3K9me3 weakened. H4K20me3 backed off. What didn’t change? The bulk histone H3 signal, suggesting nucleosomes were still there but their chemical flags had shifted. These effects showed up whether they knocked down SMARCAD1 transiently or in stable lines, arguing it wasn’t just a slow adaptation to the loss. Bring SMARCAD1 back and the system rebounded: wild-type SMARCAD1 restored H3K9me3 and KAP1 binding at IAPs and at MMERVK10C, another class II element. And there was a clear hierarchy. When they took out SETDB1, H3K9me3 predictably dropped at these retroelements—but SMARCAD1’s presence on those sites didn’t budge. Flip the experiment the other way and deplete SMARCAD1, and SETDB1 recruitment to IAPs fell, even though the total cellular level of SETDB1 stayed the same. Put simply, SMARCAD1 gets there early. It helps SETDB1 find and stay on these elements. How does SMARCAD1 know where to go? KAP1 is the guide. SMARCAD1 carries a CUE1 domain, a small module that mediates its interaction with KAP1. Mutating that domain—changes that break the SMARCAD1–KAP1 handshake—markedly reduced SMARCAD1 occupancy at retroelements. That puts KAP1 upstream in recruitment. But once SMARCAD1 is on site, the relationship becomes two-way. Lose SMARCAD1 and KAP1 binding to IAPs sags. Maximal KAP1 occupancy at these elements actually needs an intact SMARCAD1 ATPase. So, KAP1 helps bring SMARCAD1 in, and SMARCAD1, through its motor, helps KAP1 stick. That motor turned out to be the crux. The team engineered a SMARCAD1 mutant—K523R—that keeps the protein structure but knocks out its ATPase activity. In biochemical assays, that mutant could still grab KAP1. The handshake was fine. But in cells, it behaved very differently on chromatin. The ATPase-dead protein showed much weaker binding at IAPs and other class I and class II retroelements. And when expressed in cells lacking endogenous SMARCAD1, the ATPase mutant could not restore KAP1 occupancy at ERVs; in some loci, KAP1 levels dropped even further than with SMARCAD1 depletion alone. Wild-type SMARCAD1 reversed the loss. Even a SMARCAD1 variant that was defective in binding KAP1 but still had an active ATPase brought KAP1 occupancy close to normal. That’s a striking result. It says the remodeling activity itself—spending ATP to reposition or stabilize nucleosomes—can support KAP1 residence, even if the direct protein–protein tether is weakened. Binding is not enough. Hydrolysis matters. The methylation mark followed the same rule. H3K9me3 at IAPs fell when SMARCAD1 was lost. Put wild-type SMARCAD1 back and H3K9me3 returned. Try either the ATPase-dead mutant or the CUE1 mutant, and the mark stayed low. SETDB1’s grip on IAPs also loosened when SMARCAD1 was depleted. All of this points to a simple chain of cause and effect: KAP1 recruits SMARCAD1 through CUE1; SMARCAD1 uses ATP to remodel local chromatin; that remodeling helps keep KAP1 and SETDB1 in place; SETDB1 lays down H3K9me3 and, together with H4K20me3 and HDAC1, builds a robustly repressive compartment over the retroviral sequence. What happens when that chain breaks? The retrovirus talks—and so do its neighbors. In embryonic stem cells, knocking down SMARCAD1 caused transcripts from SMARCAD1-bound ERV subfamilies to rise within days. Some elements responded earlier or more strongly than others, but the pattern was clear. And it wasn’t an artifact of cells drifting toward differentiation; pushing the cells to differentiate did not switch these ERVs on. Neighboring genes got caught in the crossfire. Loci close to derepressed retroelements—genes like the Bglap3 gene, the Prnp gene, the Cntnap3 gene, the Cml2 gene, and the Zfp575 gene—picked up expression within a few kilobases of the transposon. Around the Bglap3 gene, mRNA levels went up by about two to six fold, and enhancer-proximal assays clocked a three to five fold boost tied to the nearby IAP. Restore SMARCAD1, and these genes quieted down again. Wild-type SMARCAD1 put the lid back on both the retroviral elements and the collateral gene activation; the ATPase-dead mutant did not. Pulling the evidence together, a remodeler-anchored model of ERV silencing emerges. KRAB zinc-finger proteins and KAP1 still call the shots on which sequences get targeted. But at those sequences—especially class I VL30s and class II IAPs—SMARCAD1 arrives via its CUE1 link to KAP1 and then does the heavy lifting. Its ATPase motor stabilizes SMARCAD1’s own occupancy, shores up KAP1 retention, and promotes SETDB1-dependent H3K9me3. The result is a chromatin environment that’s not just marked as repressed but physically resistant to activation. H4K20me3 joins the scene, HDAC1 tightens the wrapping, and the entire neighborhood becomes inhospitable to transcription. It’s worth appreciating the scale of what’s being contained here. IAPs alone are abundant—on the order of one thousand to two thousand copies—and together, they cover a few percent of the mouse genome. Across mammals, endogenous retroviruses make up roughly eight to ten percent of our DNA. In that light, SMARCAD1’s role is not a niche tweak. It’s a system for guarding a significant fraction of the genome during a window when the nucleus is otherwise permissive. There are a couple of mechanistic threads that make this story especially satisfying. First, the separation of recruitment and maintenance. KAP1 directs SMARCAD1 to the right addresses via the CUE1 domain, but the long-term lease—stable KAP1 and SETDB1 occupancy—depends on ATP-driven remodeling. Second, the asymmetry with SETDB1. Take away SETDB1 and SMARCAD1 still finds ERVs; take away SMARCAD1 and SETDB1 can’t hold on. That’s a clean upstream–downstream relationship. And third, the nuance that a SMARCAD1 variant with a working ATPase but weakened KAP1 binding can still bolster KAP1 occupancy on DNA suggests that remodeling may create or preserve nucleosome configurations that favor KAP1’s residence, even when direct tethering is compromised. If you’re thinking ahead to what this means beyond mouse stem cells, you’re not alone. The logic here—sequence-specific targeting, scaffold assembly, ATP-dependent chromatin remodeling to lock in a repressive state—feels general. It raises questions. What exactly is the remodeling step? Is SMARCAD1 sliding nucleosomes to occlude retroviral promoters? Is it promoting histone exchange that favors binding by KAP1 and SETDB1? There are biochemical hints that sumoylation can tune recruitment at these sites; KAP1 is a known SUMO target, and CUE domains often read ubiquitin-like signals. It’s tempting to imagine a feedback loop where modifications on KAP1 or nearby histones aid SMARCAD1 recruitment, which then stabilizes KAP1 further. And then there’s inheritance. Embryonic stem cells divide quickly. Heterochromatin has to be reassembled after each S phase. A remodeler with a built-in motor is exactly the kind of factor you’d want to help re-establish repressive chromatin over retroelements after replication forks plow through. Watching how SMARCAD1 tracks with the replication machinery at ERVs, or mapping nucleosome positions before and after ATPase disruption, would test that idea directly. For now, the key takeaway from Sachs and colleagues is crisp. In the canonical KRAB–KAP1–SETDB1 pathway that keeps endogenous retroviruses quiet, ATP-dependent remodeling is not an optional accessory. It’s a required step. SMARCAD1—through its CUE1-guided recruitment and its ATPase-driven action—stabilizes the very platform that deposits and maintains H3K9me3. Take away the motor and the marks fade, KAP1 slips, retroviruses speak, and nearby genes get swept up in the noise. In a genome where ancient viruses make up a surprisingly large share of the landscape, that motor is not just moving nucleosomes. It’s keeping the past from interrupting the present.