Summary: Researchers have found that the three-dimensional folding of the genome is substantially altered in specific brain cells of people with Alzheimer’s disease, revealing a new layer of molecular pathology linked to disrupted gene regulation and tissue organization in the prefrontal cortex.
The study combines single-cell GAGE-seq multiomics, spatial tissue mapping, and a transformer-based deep learning model called Hicformer to connect 3D genome architecture directly to gene expression and disease-related tissue changes. The authors report a characteristic loss of clear separation between active and inactive genomic domains — a phenomenon they describe as “increased compartment mingling.”
This structural breakdown diminishes regulatory contacts between genes and enhancers, contributing to reduced synaptic gene programs in neurons, metabolic and stress responses, and senescence-associated programs in microglia. The work positions higher-order chromatin reorganization alongside amyloid-beta and tau pathology as an important regulatory layer in Alzheimer’s disease.
Key Facts
- Increased compartment mingling: In Alzheimer’s brain cells, large active (A) and inactive (B) chromatin domains lose their spatial separation, correlating with lower overall gene activity.
- Weakened regulatory connections: Promoter-proximal gene-enhancer contacts are weakened, while abnormal midrange and long-range genomic interactions increase across multiple cell types.
- Hicformer AI framework: The team developed Hicformer, a transformer-based deep learning model that integrates DNA sequence, large-scale folding features, and local 3D contact maps to predict cell-type-specific gene activity and prioritize regulatory elements.
- Direct link to cellular pathology: 3D genome decay aligns with impaired synaptic and neuronal programs, altered metabolic and stress pathways, and activation of senescence-related programs in microglia.
- Multi-scale spatial integration: Using single-cell GAGE-seq together with spatial transcriptomics allowed mapping of genome folding changes onto intact prefrontal cortex tissue, connecting molecular alterations to tissue-level organization.
Source: University of Pittsburgh
Researchers from Carnegie Mellon University’s School of Computer Science, the University of Pittsburgh School of Medicine, and the University of Washington report findings that highlight a previously underexplored dimension of Alzheimer’s disease biology and point to new avenues for mechanistic studies and therapeutic exploration.
The investigators measured gene expression and chromosomal contacts in the same single cells using GAGE-seq, then placed those observations in tissue context with spatial transcriptomics of postmortem prefrontal cortex samples from donors with and without Alzheimer’s disease. Integrating these datasets enabled a direct link between genome folding, gene regulation, and how affected cells are arranged within cortical tissue.
“Alzheimer’s disease cannot be understood one layer at a time,” said Jian Ma, Ray and Stephanie Lane Professor of Computational Biology and senior author on the study. “The genome’s 3D structure is a fundamental regulatory layer that links DNA sequence to gene activity. By integrating genome folding, cell state, and tissue context, we can move beyond cataloguing changes toward understanding which mechanisms to test next.”
To assemble this multi-scale picture, the team analyzed prefrontal cortex tissue from participants in a long-term dementia study who donated their brains after death. GAGE-seq provided paired measurements of gene expression and 3D genome contacts in individual cells. Those data were integrated with chromatin accessibility maps and spatial transcriptomics from adjacent tissue sections, allowing the researchers to relate structural genome features to regulatory activity and to visualize where altered cells and programs reside in the cortex.
A central computational advance was Hicformer, a transformer-based AI model that combines DNA sequence information, broad genome-folding signals, and local contact maps to predict gene activity across cell types. Hicformer served as a computational testbed for assessing when 3D genome structure contributes explanatory power beyond DNA sequence alone and for prioritizing distal regulatory elements whose effects are mediated by chromatin contacts.
Across major brain cell types, the study found a reproducible shift in contact patterns in Alzheimer’s disease: fewer short-range interactions and a relative increase in mid-to-long-range contacts. Although A/B compartment identities were broadly preserved, their spatial segregation weakened, producing the increased compartment mingling signature associated with reduced gene expression. Promoter-centered contacts with nearby regulatory elements tended to weaken, while some midrange interactions grew stronger, particularly at sites linked to loop organization.
These architectural changes track with cell type–specific transcriptional remodeling: neurons showed reductions in synaptic and neurotransmission programs, glial cells exhibited metabolic and stress-related alterations, and microglia displayed senescence-associated activation. Mapping these molecular signatures onto tissue architecture revealed disrupted cell neighborhoods and altered spatial coordination of gene programs in disease-affected cortex.
By demonstrating that 3D genome features add predictive value for disease-related gene expression beyond sequence alone, the work provides a prioritized list of regulatory elements and loci for follow-up functional studies. The authors propose that genome folding should be considered a key regulatory layer in Alzheimer’s disease and a potential source of new therapeutic targets.
Funding: This research was supported by grants from the National Institutes of Health.
Contributing authors from Carnegie Mellon included doctoral students Shahul Alam and Shike Wang and postdoctoral research associate Junjie Tang. Additional collaborators from the University of Pittsburgh included doctoral students Alexander K. Kunisky and Jude Baroudi, post-baccalaureate research fellows Sahar and Sahel Ghorbanikalateh, and visiting scholar Shihan Wang. The team also included investigators from the Broad Institute of MIT and Harvard, UCLA, the University of Washington, and the Rush Alzheimer’s Disease Center.
Key Questions Answered:
Q: What is “increased compartment mingling” in the context of Alzheimer’s disease?
A: In healthy nuclei, active (A) and inactive (B) regions of the genome are spatially separated. In Alzheimer’s brain cells, these boundaries become less distinct, allowing active and inactive DNA regions to mix physically. This mixing is associated with suppressed gene expression across affected cell types.
Q: What is Hicformer and how was it used?
A: Hicformer is a deep learning transformer model developed by the research team. It integrates DNA sequence, large-scale genome-folding features, and local 3D contact maps to predict cell-type-specific gene activity. The model helps evaluate when structural genome information is necessary to explain disease-related transcriptional changes and to prioritize regulatory regions for further study.
Q: Why study 3D genome folding if amyloid plaques and tau tangles are already known hallmarks?
A: Amyloid and tau are important pathological hallmarks but do not fully explain how specific cellular programs fail in Alzheimer’s. The 3D genome provides an additional regulatory layer that influences whether genes are turned on or off. Mapping genome folding and its disruption can reveal mechanisms that drive cellular dysfunction and point to new intervention strategies aimed at preserving neural circuits.
Editorial Notes:
- This article was edited by a Neuroscience News editor.
- The cited journal paper was reviewed in full.
- Additional context was added by staff.
About this Alzheimer’s disease and genetics research news
Author: Allison Hydzik
Source: University of Pittsburgh
Contact: Allison Hydzik – University of Pittsburgh
Image: The image is credited to Neuroscience News
Original Research: Open access. “Single-cell multiomics connects 3D genome and transcriptome alterations in Alzheimer’s disease” by Yang Zhang, Xinyue Lu, Alexander K. Kunisky, Shahul Alam, Junjie Tang, Ruochi Zhang, Shike Wang, Han Zhang, Jude Baroudi, Walid Ichcho, Deyong Jia, Sahar Ghorbanikalateh, Sahel Ghorbanikalateh, Shihan Wang, David A. Bennett, Hansruedi Mathys, Zhijun Duan, Jian Ma. DOI: 10.1126/science.adz1652
Abstract
Single-cell multiomics connects 3D genome and transcriptome alterations in Alzheimer’s disease
INTRODUCTION
Alzheimer’s disease is the leading cause of dementia, characterized by progressive cognitive decline. While many studies have cataloged transcriptional changes across brain cell types, the molecular mechanisms driving those changes remain incompletely understood. Gene regulation depends not only on DNA sequence and epigenetic marks but also on the three-dimensional folding of the genome within the nucleus. How 3D genome organization changes in Alzheimer’s and how these changes relate to cell type–specific gene dysregulation in the human brain have been largely unexplored.
RATIONALE
The study aimed to determine whether alterations in 3D genome folding are linked to the gene expression programs disrupted in Alzheimer’s disease and whether such links can be detected at single-cell resolution in human brain tissue. To address this, the researchers applied GAGE-seq to measure gene expression and chromosomal contacts in the same cells, integrated chromatin accessibility data, and mapped these molecular features onto spatial transcriptomic profiles of intact tissue. They also developed a transformer-based predictive model to test when 3D genome structure is necessary to explain disease-linked expression changes.
RESULTS
Across major brain cell types, Alzheimer’s disease was associated with a consistent shift in genome contact patterns: fewer short-range interactions and more mid-to-long-range interactions. Although compartment identities were broadly preserved, active and inactive regions exhibited increased mixing, indicating weakened compartment segregation. These structural changes corresponded to broad, cell type–specific transcriptional remodeling affecting disease-relevant pathways and aligned with existing clinical target landscapes.
Promoter-proximal interactions at accessible regulatory elements tended to weaken, while some midrange interactions became more prominent, particularly at loci involved in loop organization. The study also found AD-related activation of senescence programs in microglia and sex-dependent dysregulation of X-linked genes in females, accompanied by corresponding 3D genome changes at implicated loci.
Hicformer demonstrated that 3D genome features provide explanatory power beyond DNA sequence alone for AD-relevant gene expression, enabling prioritization of distal regulatory elements whose influence is mediated through chromatin contacts. Integrating these molecular features with spatial transcriptomics placed them within tissue context and revealed altered cell neighborhoods and disrupted spatial coordination of gene programs in Alzheimer’s disease.
CONCLUSION
This study provides a multiscale map linking 3D genome remodeling to cell type–specific gene expression changes and to spatial tissue reorganization in Alzheimer’s disease. By establishing genome folding as a key regulatory layer associated with AD pathology, the work offers a framework and resource for generating mechanistic hypotheses, prioritizing regulatory elements, and selecting gene programs for future functional testing and therapeutic investigation.