Shared and organ-specific gene expression programs of fibrotic diseases

This study presents a large-scale, cross-organ single-cell transcriptomic meta-analysis of over five million cells from human heart, liver, kidney, and lung tissues to construct a unified fibrosis atlas that identifies both shared and organ-specific gene expression programs, ultimately providing an open resource to accelerate the discovery of antifibrotic therapies.

The Gist

Imagine your body as a bustling city. When a neighborhood gets damaged—say, by a fire (infection), a storm (injury), or a toxic spill (chemical exposure)—the city sends in a construction crew to fix it. This crew is made of fibroblasts, the body's natural repair workers. Their job is to lay down "scaffolding" (collagen and other proteins) to patch the hole.

Usually, once the patch is done, the crew packs up and leaves. But in fibrosis, the construction crew gets stuck in "overdrive." They keep laying down scaffolding long after the damage is fixed, eventually turning a flexible neighborhood into a rigid, scarred wasteland. This happens in the heart, lungs, kidneys, and liver, often leading to organ failure.

For a long time, scientists studied these "construction crews" in isolation. They looked at the lung crew, then the kidney crew, then the heart crew, as if they were completely different species. This made it hard to find a universal cure because we didn't know if they were all following the same bad instructions.

This paper is like a massive, global summit where all these construction crews finally meet to compare notes.

Here is what the researchers did, explained simply:

1. The Great Data Merge (The "City Council" Meeting)

The researchers gathered data from 20 different studies involving over 5 million cells from human hearts, lungs, kidneys, and livers. Imagine taking thousands of individual reports from construction sites across the world and merging them into one giant, searchable database. They looked at both healthy tissues and scarred (fibrotic) tissues.

2. Finding the Common Language (Shared vs. Local Dialects)

When they compared the data, they found two fascinating things:

• The Local Dialects: Some genes (instructions) were unique to specific organs. For example, the "kidney crew" had some specific tools that the "lung crew" didn't use. This explains why a drug that works for the liver might not work for the heart.
• The Universal Language: Despite the differences, the crews were speaking the same "bad language" when it came to scarring. They found a shared set of instructions that all these crews were following to create excessive scars. It's like realizing that construction crews in Tokyo, New York, and Berlin are all using the same faulty blueprint that causes them to build too much wall.

3. The "Super-Scarring" Crew (Disease Fibroblasts)

Within the construction crews, the researchers identified a specific subgroup of workers—the "Disease Fibroblasts." These are the troublemakers.

• In healthy tissue, these workers are quiet.
• In fibrotic tissue, they wake up and start shouting orders to build more scaffolding.
• The amazing discovery? These "troublemaker" workers look and act almost identically whether they are in the heart, lung, kidney, or liver. They are essentially the same bad actor wearing different uniforms.

4. The Communication Network (Who is Talking to Whom?)

Construction crews don't work alone; they talk to each other. The researchers mapped out who was talking to whom.

• They found that in fibrotic tissues, the cells are constantly sending "build more!" signals to each other.
• They identified specific "radio frequencies" (molecular signals) that are turned up loud in all four organs. One of the loudest signals came from a molecule called Tenascin-C. It's like a siren that tells the whole city to stop healing and start scarring.

5. The Treasure Map (New Targets for Cures)

Because they looked at all the organs together, the researchers found some "hidden gems" that previous studies missed.

• They found genes like MOXD1 and FGF14 that were previously only known to be involved in kidney or lung scarring. But this study showed they are also active in the heart and liver!
• This is like realizing that a tool you thought was only for fixing roofs is actually the key to fixing the foundation, the plumbing, and the electrical wiring too.

Why This Matters

Currently, we have very few drugs to treat fibrosis, and most only work for one specific organ (like the lungs). This is like having a key that only opens the front door of a house, but not the back door or the basement.

This paper provides a master key. By showing that the "bad construction crew" is fundamentally the same across different organs, it suggests that we might be able to develop one drug that stops the scarring process in the heart, lungs, kidneys, and liver all at once.

In short: The researchers took a chaotic pile of puzzle pieces from four different cities, put them together, and realized they were all part of the same picture. Now, instead of trying to solve four different puzzles, we can finally focus on solving the one big puzzle of fibrosis.

Technical Summary

1. Problem Statement

Fibrosis is a progressive, irreversible condition characterized by excessive extracellular matrix (ECM) deposition, leading to organ dysfunction and failure. While fibrotic diseases affect nearly all organs (heart, liver, kidney, lung), current research largely treats them in isolation.

• The Gap: Proposed shared mechanisms are often based on indirect comparisons of independent datasets rather than unified, systematic cross-organ meta-analyses.
• The Challenge: Existing studies focus on specific disease etiologies within single tissues, making it difficult to distinguish between organ-specific adaptations and conserved, universal fibrotic mechanisms. Furthermore, the lack of approved anti-fibrotic drugs for multiple tissues suggests a need to identify truly conserved therapeutic targets.
• Objective: To construct a unified, large-scale single-cell atlas of human fibrosis to systematically identify both shared (cross-organ) and organ-specific transcriptomic programs, cell states, and intercellular communication patterns.

2. Methodology

The authors performed a comprehensive meta-analysis integrating single-cell (scRNA-seq) and single-nucleus (snRNA-seq) transcriptomic data.

Data Curation & Integration:

• Scope: Curated 20 studies comprising 5.1 million cells from 953 human samples across four major organs: Heart, Lung, Kidney, and Liver.
• Diversity: Covered >25 disease etiologies (e.g., IPF, DCM, DKD, MASH).
• Harmonization:
  • Cells were grouped into five broad lineages: Endothelial, Epithelial, Mesenchymal, Lymphoid, and Myeloid.
  • Cell type labels were harmonized; for datasets lacking clear annotations, scVI/scANVI was used for label transfer.
  • Technical variability (e.g., sequencing depth, mitochondrial content) was assessed and accounted for.

Analytical Framework:

1. Quality Control & Filtering: Studies with negligible molecular differences between fibrotic and reference groups (assessed via ScDist) were excluded.
2. Compositional Analysis: Linear mixed models were used to quantify changes in cell type proportions between fibrotic and healthy states.
3. Multicellular Factor Analysis: MOFA2 was employed to extract latent factors representing shared molecular programs across cell types within an organ.
4. Meta-Analysis of Differential Expression:
   • Level 1: Differential expression (DE) calculated per study using PyDESeq2.
   • Level 2 (Organ): Random-effects meta-analysis (inverse-variance weighting) to derive Organ-Consensus Genes.
   • Level 3 (Cross-Organ): Second meta-analysis to derive Cross-Organ Consensus Genes.
5. Predictive Modeling: Disease score prediction models (using top DEGs from one study to predict status in another) were used to quantify the conservation of transcriptional programs (AUROC metrics).
6. Cell-Cell Communication:
   • Transcriptomic: LIANA consensus resource used to infer ligand-receptor (LR) interactions.
   • Spatial: Integrated with Visium spatial transcriptomics datasets (4 organs) to calculate bivariate Moran's R for spatial co-localization of LR pairs, specifically focusing on scar-localized signals (using COL1A1 as a scar anchor).
7. Functional Interpretation: Decoupler and ULM enrichment analyses were used to infer Transcription Factor (TF) activities and Gene Ontology (GO) pathways.

3. Key Contributions

• The Organ Fibrosis Atlas: Creation of the largest cross-organ single-cell fibrosis resource to date, publicly available at organfibrosis.saezlab.org.
• Unified Meta-Analysis Pipeline: A robust statistical framework distinguishing between organ-specific noise and conserved biological signals across heterogeneous datasets.
• Definition of Disease Fibroblasts: Identification of a specific "disease-associated fibroblast" subpopulation across organs that converges on a common gene profile, distinct from general mesenchymal cells.
• Spatial-Transcriptomic Integration: Novel methodology combining scRNA-seq meta-analysis with spatial co-localization to pinpoint therapeutic targets specifically within fibrotic scar regions.

4. Key Results

A. Cellular Composition & Shared Signals:

• Proportion Changes: Fibrosis increased mesenchymal cells in the lung, but lymphoid cells in the heart and kidney. Liver studies showed high variability, likely due to the organ's regenerative capacity and diverse fibrotic phenotypes.
• Conservation: Heart, Lung, and Kidney showed strong shared transcriptional signatures (high AUROC in cross-study prediction). Liver showed the lowest agreement.
• Cell Lineage Specificity: Mesenchymal and Endothelial cells displayed the highest degree of shared gene regulation across organs. Lymphoid and Myeloid cells showed high organ-specificity.

B. Gene Expression Programs:

• Organ-Consensus Genes: Identified hundreds of upregulated genes per organ (e.g., SEZ6L in heart mesenchymal cells; POSTN in lung).
• Cross-Organ Consensus: While most genes were organ-specific, a core set of genes was consistently regulated across tissues.
  • Top Upregulated: LAMB1, PCDH17, ZNF521 (Endothelial); SMOC2, LOXL1, MOXD1, FGF14 (Mesenchymal).
  • Top Downregulated: PPARG (a known anti-fibrotic factor).
• Transcription Factors: Conserved activation of pro-fibrotic TFs (e.g., TWIST1, SP3, SCX, SMAD3) and reactivation of developmental programs (e.g., HOX genes, morphogenesis pathways).

C. Disease Fibroblasts:

• A specific cluster of "disease-associated fibroblasts" was identified in all organs.
• These cells showed reduced organ-specificity in their gene programs compared to general mesenchymal cells, converging on a universal fibrotic signature.
• Functional Shift: Upregulation of ECM assembly, collagen fibril organization, and secretory pathways; downregulation of oxidative phosphorylation and ion transport.

D. Cell-Cell Communication & Scar Targets:

• General Communication: Dominated by ECM components (Collagens, Fibronectin) interacting with Integrins.
• Scar-Specific Targets: By filtering for genes spatially co-localized with COL1A1 (scar marker) and excluding structural collagens, the authors identified high-confidence therapeutic targets:
  • Top Hits: TIMP1 (Metalloproteinase inhibitor 1), THBS2 (Thrombospondin 2), POSTN (Periostin), and Tenascin C (TNC).
  • Significance: TNC inhibition showed efficacy in systemic sclerosis models; this study suggests its potential as a broad-spectrum anti-fibrotic target across heart, lung, kidney, and liver.

5. Significance

• Therapeutic Potential: The study validates the hypothesis that targeting conserved molecular mechanisms (e.g., MOXD1, TNC, PPARG agonists) could yield broadly effective antifibrotic drugs, moving beyond the current limitation of organ-specific treatments.
• Resource Accessibility: The interactive web resource allows researchers to query specific genes, cell types, and disease etiologies, accelerating hypothesis generation.
• Methodological Advance: Demonstrates the power of integrating single-cell meta-analysis with spatial transcriptomics to move from "differential expression" to "spatially relevant therapeutic targets."
• Biological Insight: Confirms that while fibrosis manifests differently across organs, the core molecular engine—driven by specific mesenchymal and endothelial states and developmental reprogramming—is highly conserved.

Limitations Noted: Technical heterogeneity between datasets (snRNA vs. scRNA), limited patient metadata (severity, exact location), and the aggregation of diverse disease etiologies under a single "fibrosis" label may obscure subtle condition-specific nuances.