Altered chromatin accessibility and nucleosome positioning landscape upon HDAC and LSD1 inhibition in cancer cell

This study employs a multi-modal deep learning platform to demonstrate that dual inhibition of HDAC and LSD1 enzymes in cancer cells alters chromatin accessibility and nucleosome positioning, revealing a JunB-mediated mechanism that displaces the CoREST-RUNX complex to trigger apoptosis and suggesting optimized combinatorial epigenetic therapies.

The Gist

The Big Picture: Unlocking the Cell's "Locked" Rooms

Imagine your DNA is a massive library inside a cell. In a healthy cell, the books (genes) are organized, and the librarians know exactly which books to open for reading and which to keep locked away.

In cancer, the library gets messy. The "locks" on the books get broken or stuck, causing the cell to read the wrong books (turning on bad genes) or ignore the good ones (turning off tumor suppressors).

Two types of "librarians" are responsible for this mess:

1. The "Lock-Keepers" (HDACs): They tighten the locks, making DNA hard to read.
2. The "Lock-Removers" (LSD1s): They usually take locks off, but in cancer, they sometimes take them off the wrong books.

The Goal: Scientists want to use drugs to fix these locks. But should they use one drug to fix one type of lock, or two drugs to fix both at once? This paper tries to answer that question.

---

The New Tool: A "Smart Camera" for DNA

To study this, the researchers invented a new super-tool called NicE-viewSeq and a computer brain called NicEL.

• The Old Way: Usually, scientists have to smash the cell open to read the DNA, losing the picture of where everything was.
• The New Way (NicE-viewSeq): Imagine taking a high-resolution photo of the library without breaking it open. They use special glowing dyes that stick only to the "open" books (accessible DNA).
• The Computer Brain (NicEL): They built an AI (like a smart camera app) that looks at these photos, finds every single cell nucleus, and calculates exactly how "open" the DNA is. It's like having a robot that counts how many books are open on every shelf in the library instantly.

The Experiment: One Drug vs. Two Drugs

The researchers tested three scenarios on cancer cells (HT1080 cells):

1. Drug A (HDAC Inhibitor): Loosens the tight locks.
2. Drug B (LSD1 Inhibitor): Stops the wrong locks from being removed.
3. The Combo (Both Drugs): Doing both at the same time.

The Results:

• Drug A alone: Opened up some books, but mostly in the middle of the "chapters" (gene bodies).
• Drug B alone: Opened up books near the "start" of the chapters (promoters).
• The Combo: This was the winner. It didn't just open a few more books; it threw the doors wide open across the entire library. The DNA became much more accessible, and the "spreading" of open regions was massive.

The Mechanism: The "Musical Chairs" of Proteins

Here is the most interesting part: How did the combo drug work so well?

Imagine the DNA is a dance floor.

• The "Bad Dancers" (CoREST/RUNX): In the cancer cell, a group of proteins called CoREST and RUNX are dancing on the floor, blocking the music (gene expression) and keeping the cell stuck in a cancerous state.
• The "Good Dancers" (JunB): When the researchers gave the drugs, they kicked the "Bad Dancers" off the floor.
• The Switch: In their place, a new protein called JunB (a member of the bZIP family) jumped onto the dance floor.

The Analogy: Think of the DNA as a stage. The "Bad Dancers" (CoREST/RUNX) were standing in the spotlight, blocking the show. The drugs acted like a stagehand who swept them away and replaced them with "Good Dancers" (JunB). JunB then started a new show that told the cancer cell to stop growing and start dying (apoptosis).

Why This Matters

1. Synergy is Key: Using two drugs together wasn't just "1 + 1 = 2." It was "1 + 1 = 10." The combination created a unique environment that neither drug could achieve alone.
2. The "Switch" Mechanism: The study found a specific "switch" (JunB replacing CoREST/RUNX) that triggers the cancer cell's self-destruct button. This gives doctors a new target to aim for.
3. Better Tools: The new AI tool (NicEL) they built can help doctors test these drugs faster and more accurately in the future, potentially leading to better cancer treatments with fewer side effects.

The Bottom Line

This paper shows that by using a clever combination of two epigenetic drugs, we can force cancer cells to "rearrange their furniture." We kick out the bad proteins that keep the cancer alive and replace them with proteins that tell the cell to die. It's like fixing a broken library by not just unlocking the doors, but swapping out the librarians for ones who actually want to help.

Technical Summary

1. Problem Statement

Epigenetic enzymes (writers, readers, and erasers) regulate chromatin landscapes and contribute to tumor heterogeneity. While therapeutic targeting of these enzymes (e.g., HDAC inhibitors and LSD1 inhibitors) shows clinical promise, the comparative efficacy of mono-inhibitor versus dual-inhibitor strategies remains unclear. Specifically, there is a lack of mechanistic understanding regarding:

• How mono- vs. dual-inhibition differentially alters chromatin accessibility and nucleosome positioning.
• The specific transcription factor (TF) dynamics and regulatory axes triggered by these treatments.
• The ability to quantitatively assess these changes at the single-cell level while preserving spatial context, as existing methods (like standard ATAC-seq) struggle with chemically fixed archival samples and lack spatial resolution.

2. Methodology

The authors developed a multi-modal platform integrating deep learning, spatial imaging, and high-throughput sequencing to analyze chromatin dynamics in HT1080 cancer cells.

• Experimental Design:
  • Cell Model: HT1080 fibrosarcoma cells.
  • Treatments:
    • Mono-inhibition: LSD1 inhibitor (GSK-2879552) and HDAC inhibitor (Romidepsin).
    • Dual-inhibition: Sequential treatment with GSK-2879552 followed by Romidepsin.
    • Controls: DMSO and various other epigenetic inhibitors (JQ-1, PFI-2, WDR5, UNC0379) for model validation.
• NicE-viewSeq (Imaging & Spatial Profiling):
  • A method using formaldehyde-fixed cells labeled with fluorescent dNTPs (for accessible chromatin) and biotinylated dATP (for sequencing).
  • Enables simultaneous microscopic imaging of chromatin accessibility and high-throughput sequencing.
• NicEL (NicE-view Learning):
  • A custom deep learning pipeline developed by the authors.
  • Stage 1: Nuclei detection using a StarDist framework with a U-Net backbone to segment densely packed nuclei.
  • Stage 2: Calculation of the Accessible Chromatin Index (ACI), a normalized metric (ratio of accessibility signal to DNA label signal) to quantify chromatin accessibility per nucleus, correcting for imaging variability.
• Genomic Sequencing & Analysis:
  • NicE-seq: High-throughput sequencing of biotinylated accessible DNA to map genome-wide accessibility.
  • NEED-seq: Epitope-targeted sequencing to map histone marks (H3, H3K4me1/2, H3K9ac) and protein binding (RCOR1, LSD1, HDAC1, RUNX, JunB).
  • RNA-seq: To correlate chromatin changes with gene expression.
  • Bioinformatics: Differential peak analysis (DiffBind), motif enrichment (Homer), nucleosome repositioning (DANPOS3), and TF interaction networks (STRING/Rosetta2Fold/Leiden clustering).
  • Validation: Western blotting and gel filtration chromatography.

3. Key Contributions

1. Development of NicEL: A robust, automated deep learning pipeline that quantifies chromatin accessibility from NicE-view images with high accuracy and reduced variability compared to manual analysis, enabling single-nucleus resolution in fixed samples.
2. Comparative Epigenetic Profiling: A comprehensive comparison of mono- vs. dual-inhibition strategies, revealing that dual inhibition induces a synergistic, genome-wide expansion of accessible chromatin regions.
3. Discovery of the CoREST-RUNX-JunB Axis: Identification of a novel mechanistic switch where the JunB transcription factor (bZIP family) displaces the CoREST-RUNX complex at specific chromatin regions upon inhibitor treatment, triggering apoptotic pathways.
4. Mechanistic Link to Nucleosome Repositioning: Demonstrating that histone spreading and nucleosome repositioning (specifically shifts of 10–75 bp) are primary drivers of altered transcriptional programs and transposable element (TE) activation.

4. Key Results

• Chromatin Accessibility Trends:

  • Dual Inhibition Synergy: Dual treatment (LSD1i + HDACi) resulted in a ~2x increase in ACI scores compared to controls, significantly higher than mono-treatments (~0.3–0.5x).
  • Genomic Distribution: Dual inhibition caused a massive spreading of accessible chromatin (up to 3.5–4 kb), particularly at Transcription Start Sites (TSS). Mono-HDAC inhibition showed a more diffused pattern, while mono-LSD1 inhibition was TSS-specific.
  • Peak Dynamics: While dual inhibition increased the width of accessible regions, it paradoxically reduced the number of distinct peaks compared to HDACi alone, suggesting a merging of accessible domains.

• Histone Marks and Nucleosome Repositioning:

  • Inhibition led to a global spreading of H3K4me1/2 and H3K9ac marks.
  • Nucleosome Shifting: Approximately 15–20% of nucleosomes shifted position by 10–75 bp, with a small fraction shifting >75 bp. This repositioning correlated with the expansion of accessible chromatin.

• Transcription Factor Switching:

  • CoREST Complex Disruption: Inhibition reduced the correlation between CoREST components (RCOR1, LSD1, HDAC1).
  • TF Switch: In control cells, RUNX (Runt family) co-localized with the CoREST complex. Upon treatment, JunB (bZIP family) levels increased and displaced RUNX/CoREST at accessible regions.
  • Network Analysis: TF interaction networks showed that ETS, RUNT, and bZIP families were central seed nodes. Dual inhibition maximized TF community variability.

• Transcriptional Outcomes:

  • Gene Expression: Dual inhibition induced the highest number of differentially expressed genes (DEGs), with ~70% upregulated.
  • Apoptosis: Upregulated genes were enriched for apoptotic pathways. Downregulated genes in HDACi treatment were linked to cell cycle arrest.
  • TE Activation: Nucleosome repositioning was linked to the activation of transposable elements (LTRs, LINEs, SINEs), a known mechanism for inducing viral mimicry and immune response.

5. Significance

• Therapeutic Optimization: The study provides a mechanistic rationale for using dual HDAC/LSD1 inhibitors (like Corin) over mono-therapies. The dual approach creates a convergent vulnerability by forcing a "TF switch" (JunB vs. RUNX) that drives apoptosis more effectively.
• Methodological Advancement: The NicEL/NicE-viewSeq platform bridges the gap between spatial imaging and genomic sequencing, offering a powerful tool for analyzing fixed clinical samples where traditional ATAC-seq fails.
• Biological Insight: The discovery of the JunB-mediated displacement of the CoREST-RUNX complex reveals a new regulatory axis in cancer epigenetics. It suggests that therapeutic efficacy may depend not just on enzyme inhibition, but on the subsequent reprogramming of the transcription factor landscape.
• Clinical Relevance: These findings support the development of combinatorial epigenetic therapies to overcome drug resistance and induce tumor cell death via apoptotic and viral mimicry pathways.