Histone readers MLLT1 and MLLT3 concentrate AID to confer locus specificity

Every time your immune system produces an antibody fine-tuned to fight an infection, it gambles with DNA. A specialized enzyme called activation-induced deaminase, or AID, deliberately mutates antibody genes inside B cells, generating the staggering diversity of antibodies that protects us from pathogens. But AID is a double-edged sword: the same mutagenic power that diversifies antibody genes can also corrupt other parts of the genome, and this off-target activity is a known driver of B cell lymphomas. For decades, immunologists have puzzled over a central question: how does AID know where to strike? Transcription is required for AID to work, yet the vast majority of transcribed genes in a B cell are never mutated. AID can be found sitting on many more genomic sites than it actually damages, meaning that mere physical presence is not enough. A new study published in Nature has now identified the molecular gatekeepers that decide where AID is truly licensed to act, and the answer involves a surprising mechanism borrowed from the physics of cellular organization.

A research team led by Javier M. Di Noia at the Institut de Recherches Cliniques de Montréal, working with colleagues including François Robert and Nicole J. Francis, set out to find what distinguishes the handful of genomic regions that AID actually mutates from the thousands it merely visits. Their search converged on two proteins with a storied history in cancer biology: MLLT1, also known as ENL, and its close relative MLLT3, also known as AF9. Both are components of the super elongation complex, a molecular machine that helps RNA polymerase II push through genes during transcription. Crucially, both proteins carry YEATS domains, which are so-called histone readers, modules that recognize and bind specific chemical tags on histone proteins, the spools around which DNA is wound. This ability to read chromatin marks positions MLLT1 and MLLT3 at active, acetylated regions of the genome, including enhancers and super-enhancers.

The team’s first and most striking finding came from removing these proteins. When they deleted both MLLT1 and MLLT3 from mouse B cells, every AID-dependent mutagenic process ground to a halt. Class-switch recombination, the process that lets antibodies change their effector function, was abolished. Somatic hypermutation, which introduces point mutations into antibody variable regions to sharpen their binding, disappeared entirely. In functional terms, the double-knockout cells looked as if AID itself had been deleted, even though the enzyme was still present and expressed at normal levels. Single knockouts produced partial defects, revealing that the two proteins act redundantly, with MLLT1 playing the dominant role in mice. The effect was not limited to engineered cell lines: conditional knockout mice lacking MLLT1 or MLLT3 in their B cells showed impaired antibody class switching, and chemical inhibitors targeting the YEATS domains of both proteins suppressed immunoglobulin diversification in both mouse and human B cells.

A critical clue came from mapping where MLLT1 and MLLT3 sit in the genome. Using chromatin immunoprecipitation sequencing with spike-in normalization, the researchers found that the genomic regions that AID mutates, in both mouse and human B cells, are precisely marked by unusually high occupancy of both proteins, a signature the authors denote MLLT1/MLLT3-high. This was not a loose correlation. When the team ranked genes by MLLT1 or MLLT3 signal and stratified them at an inflection point separating high from moderate occupancy, AID targets clustered overwhelmingly in the high-occupancy group. The pattern held across B cell lines, primary mouse B cells, and human lymphoma cell lines, and it extended to the super-enhancers that aberrant somatic hypermutation frequently strikes in diffuse large B cell lymphoma. The most frequently mutated loci in a catalogue of 610 lymphoma cases correlated with MLLT1 and MLLT3 enrichment, linking the mechanism directly to human disease.

One obvious explanation for the loss of AID activity in the knockout cells would be that MLLT1 and MLLT3 are simply needed to transcribe the target genes, and without them the antibody loci fall silent. The researchers tested this rigorously and found it was not the case. Transcriptional changes after deleting both readers were modest. Measurements of nascent transcription by TT-seq, along with analyses of RNA polymerase II pausing and elongation indices, showed that the antibody germline transcripts and AID’s own expression were largely preserved. The dramatic collapse in mutagenesis could not be accounted for by changes in how much or how fast the relevant genes were being transcribed. Whatever MLLT1 and MLLT3 were doing for AID, it was something beyond their canonical role in transcriptional elongation.

The mechanistic picture that emerged is one of local concentration rather than global recruitment. The researchers showed that MLLT1 and MLLT3 are dispensable for tethering AID to chromatin in general; even without them, AID still associates broadly with the genome. What the readers provide is local enrichment: they concentrate AID specifically in the regions just downstream of promoters, within roughly two kilobases of the transcription start site, which is exactly the window in which somatic hypermutation operates. Both proteins physically bind to AID, and chromatin immunoprecipitation of tagged AID revealed that its signal at target regions drops sharply in cells lacking both readers. In other words, AID can touch the genome anywhere, but only where MLLT1 and MLLT3 pile up does it reach the concentrations needed to inflict mutations.

The most conceptually adventurous part of the study concerns how MLLT1 achieves this concentration. Beyond its structured YEATS domain, MLLT1 carries a large intrinsically disordered region, a floppy protein segment that lacks a fixed three-dimensional shape. Such regions are hallmarks of proteins that drive phase separation, the process by which certain molecules coalesce into droplet-like condensates, concentrating specific partners much like oil droplets form in water. The team purified MLLT1 and MLLT3 proteins and showed in vitro that MLLT1 forms droplets that are sensitive to the aliphatic alcohol 1,6-hexanediol and to high salt, treatments that disrupt many condensates. When they incubated these droplets with extracts from cells expressing fluorescently tagged AID, AID preferentially partitioned into the MLLT1 condensates. Live-cell imaging and partition-coefficient measurements in cells reinforced the same conclusion: AID is drawn into MLLT1-rich condensates, and this depends on AID’s own interaction surface, since a point mutation in AID, R174E, reduced its partitioning and altered its protein-proximity network as measured by BioID proximity labeling.

The functional tests were decisive. Swapping the intrinsically disordered regions between MLLT1 and MLLT3 showed that MLLT1’s disordered region is what gives it its dominant role in promoting condensates that concentrate AID in mice. Most strikingly, when the researchers artificially fused AID directly to MLLT1 or to MLLT3, forcing the enzyme into proximity with the readers regardless of natural interactions, class-switch recombination and mutagenesis were restored in cells lacking both endogenous proteins. This rescue experiment demonstrates that spatial confinement with MLLT1 or MLLT3 is not merely correlated with AID activity but is sufficient to license it, arguing that the local concentration of AID is the rate-limiting ingredient for mutagenesis at any given locus.

The implications ripple outward in several directions. For basic immunology, the work solves a long-standing puzzle by proposing a licensing model: transcription opens the door, but MLLT1/MLLT3-high chromatin determines which rooms AID is allowed to enter and vandalize. For cancer biology, it offers a mechanistic explanation for why lymphomas repeatedly acquire aberrant mutations at super-enhancers, which are precisely the structures richest in super elongation complex components and their histone-reading modules. Intriguingly, MLLT1 and MLLT3 are themselves established oncogenic players, most famously as fusion partners of MLL in acute leukemia, and small-molecule chemical probes against their YEATS domains already exist. The study raises the possibility that such inhibitors, which suppressed antibody diversification in the current experiments, could in principle modulate AID’s off-target mutagenesis in disease contexts. At the same time, the findings suggest a caution: any therapy that disrupts these condensates might impair the antibody diversification that vaccines and immune memory depend on. What began as a question about a single mutagenic enzyme has ended up connecting antibody immunity, transcriptional condensates, and lymphoma genetics into one coherent physical model, and it is likely to keep both immunologists and cancer biologists busy for years to come.

Subject of Research: Mechanism of locus-specific targeting of the mutagenic enzyme AID by the histone readers MLLT1 and MLLT3 in B cells

Article Title: Histone readers MLLT1 and MLLT3 concentrate AID to confer locus specificity

Article References: Seija, N., Gannon, S., Häfner, K. A., Gemeinhardt, T. M., Ridani, J., Alvarez, D., Provencher, M., Subramani, P. G., Poitras, C., Piskor, E.-M., Möröy, T., Vonniessen, N., Mazer, B., Navarrette, M. A., Francis, N. J., Robert, F., & Di Noia, J. M. (2026). Histone readers MLLT1 and MLLT3 concentrate AID to confer locus specificity. Nature. https://doi.org/10.1038/s41586-026-11087-1

Image Credits: AI Generated

DOI: 10.1038/s41586-026-11087-1

Keywords: AID, MLLT1, MLLT3, somatic hypermutation, class-switch recombination, histone readers, super elongation complex, phase separation, condensates, B cell lymphoma, immunoglobulin genes, chromatin